{ "schema": "postcolor-fixed-quality-public-values-v1", "current_baseline": "PML", "candidate": "RAW_POSTCOLOR_M1", "measurement_state": "completed_with_missing_observations", "candidate_rows_recorded": 30, "candidate_quality_status_counts": { "relative_f0_mae_cents": { "not_applicable": 10, "available": 20 }, "unvoiced_miss_rate": { "not_applicable": 10, "available": 20 }, "automatic_mos": { "available": 30 }, "recognition": { "unavailable": 18, "available": 12 } }, "four_metrics_available_rows": 8, "old90_reevaluated": false, "statistical_units": { "recordings": 5, "speakers": 4, "seeds": [ 0, 20261015 ], "pitches_semitones": [ 0, -3, 3 ], "independent_units_increased_by_seeds": false, "formal_120_recording_evaluation": false }, "fixed_PML_harm_thresholds": { "automatic_mos_delta_below": -0.15, "relative_f0_MAE_delta_cents_above": 10, "unvoiced_miss_delta_pp_above": 2, "ASR_harm_threshold": null }, "limitations_ja": [ "F0は各方式自身のnative出力相対であり、native絶対F0真値を測らない。有声集合・分母が方式間で異なり得る。", "18候補条件は公式部分GT未整備。全文GT2録音の12条件だけASR評価。両原音CERは0で0.20ゲート通過。", "予測MOSは人間評定ではない。5録音4話者の限定観測。総合順位・採用判断は行わない。", "音響・周期観測の害をこれらの品質値で取り消さない。" ], "candidate_rows": [ { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "3fb4ef0208790e1d9a08e144860ff35699c058e53a100462576d2842e37c965f", "native_reference_sha256": null, "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3087401390075684 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "e24a1c39383ef68d78409dbc3d72621b1eca9a8b62571819aaaa9f2256748bc8", "native_reference_sha256": null, "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4289028644561768 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "56c62c71aeb2bf93cb796f11c244efdc3e993d6611656898be8e52c46b544aa4", "native_reference_sha256": "3fb4ef0208790e1d9a08e144860ff35699c058e53a100462576d2842e37c965f", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4846174172578417, "expected_voiced_frames": 158, "missing_voiced_frames": 39, "relative_error": { "count": 119, "mean": 0.17562147657614474, "mae": 2.4846174172578417, "rmse": 3.8578352789581687, "abs_p95": 9.680797128262629, "abs_max": 14.308656604444584 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2468354430379747, "expected_voiced_frames": 158, "missing_voiced_frames": 39 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5018770694732666 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "1adec25b804af2d2aa5e04663d442244cc26f6381b9c3d50a25848ef7669092c", "native_reference_sha256": "e24a1c39383ef68d78409dbc3d72621b1eca9a8b62571819aaaa9f2256748bc8", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.45458032270452, "expected_voiced_frames": 155, "missing_voiced_frames": 34, "relative_error": { "count": 121, "mean": 0.10793577528648683, "mae": 2.45458032270452, "rmse": 3.8142074679518854, "abs_p95": 9.485884691206852, "abs_max": 14.297578540668852 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.21935483870967742, "expected_voiced_frames": 155, "missing_voiced_frames": 34 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.551800012588501 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "5a237dceec200dc9344c94fa85baf7d0b853f5160902a9b00634b4f1ce76f948", "native_reference_sha256": "3fb4ef0208790e1d9a08e144860ff35699c058e53a100462576d2842e37c965f", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.80379499561333, "expected_voiced_frames": 158, "missing_voiced_frames": 11, "relative_error": { "count": 147, "mean": 0.20335123158481344, "mae": 2.80379499561333, "rmse": 4.445086561799716, "abs_p95": 9.83624351012342, "abs_max": 20.297750993054887 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.06962025316455696, "expected_voiced_frames": 158, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.343353748321533 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "60319a6cc33fa21fc4c1ce28cc426593207331e54b596f15873669c7ad36691d", "native_reference_sha256": "e24a1c39383ef68d78409dbc3d72621b1eca9a8b62571819aaaa9f2256748bc8", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.814166483934752, "expected_voiced_frames": 155, "missing_voiced_frames": 8, "relative_error": { "count": 147, "mean": 0.015511484239801313, "mae": 2.814166483934752, "rmse": 4.565950325064086, "abs_p95": 10.371135851924, "abs_max": 21.836592528378745 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.05161290322580645, "expected_voiced_frames": 155, "missing_voiced_frames": 8 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.217778205871582 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "35d484a333cf1dab94d3febcb9d0d2105a85804fe4db39537a8f57a5e5843fb3", "native_reference_sha256": null, "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.449693441390991 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "e246a24366605e5de6ce6d89e9174673cfd138c8d278c74cb046ac0049b23348", "native_reference_sha256": null, "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2923033237457275 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "8056eedc68e2ad58640646f77a5c4df4ce9facdfefb2693013f1a48ef4a760bd", "native_reference_sha256": "35d484a333cf1dab94d3febcb9d0d2105a85804fe4db39537a8f57a5e5843fb3", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.1716787977691494, "expected_voiced_frames": 141, "missing_voiced_frames": 31, "relative_error": { "count": 110, "mean": 0.35617084077631045, "mae": 2.1716787977691494, "rmse": 3.1519625445823114, "abs_p95": 5.637921583963565, "abs_max": 14.344185642064348 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2198581560283688, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.775503158569336 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "bf380796a37f19993641182d099ce967d53b205d3d64c13344855ed440f83d0a", "native_reference_sha256": "e246a24366605e5de6ce6d89e9174673cfd138c8d278c74cb046ac0049b23348", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.423271972110417, "expected_voiced_frames": 143, "missing_voiced_frames": 32, "relative_error": { "count": 111, "mean": 0.1180864208272022, "mae": 2.423271972110417, "rmse": 3.8871121658648, "abs_p95": 6.304539241379302, "abs_max": 24.770679570810216 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.22377622377622378, "expected_voiced_frames": 143, "missing_voiced_frames": 32 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.8541226387023926 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "69a04884103d8af8de3dbe10aac65d1c74e7af005e709d5c438ed16a16010549", "native_reference_sha256": "35d484a333cf1dab94d3febcb9d0d2105a85804fe4db39537a8f57a5e5843fb3", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.052278100448317, "expected_voiced_frames": 141, "missing_voiced_frames": 5, "relative_error": { "count": 136, "mean": 0.5683456019722232, "mae": 3.052278100448317, "rmse": 5.037951338699477, "abs_p95": 10.399707479053465, "abs_max": 27.76638432928407 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03546099290780142, "expected_voiced_frames": 141, "missing_voiced_frames": 5 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.459120273590088 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "c1aa72cdd4b91ad41084a4b253e11bd6f9fc9d86f47d044b9609943ff90929a0", "native_reference_sha256": "e246a24366605e5de6ce6d89e9174673cfd138c8d278c74cb046ac0049b23348", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.27440482768568, "expected_voiced_frames": 143, "missing_voiced_frames": 7, "relative_error": { "count": 136, "mean": 0.5931982188466154, "mae": 3.27440482768568, "rmse": 5.409361591908599, "abs_p95": 11.024055167283903, "abs_max": 27.743546431109557 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04895104895104895, "expected_voiced_frames": 143, "missing_voiced_frames": 7 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4007201194763184 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "6a130384d4ef170fe3271ba22740fa9fcf3f06937a9db63a274cd4a4d5157656", "native_reference_sha256": null, "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2409791946411133 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "7bff9aade99e782ef3ba9bdf6b65a01d6dabcebd2a3402c8e58868c4996d14dc", "native_reference_sha256": null, "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.038602828979492 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "6d4606089c4c9428cebe258d7036c74c04e13b155312e7bc4e5c03fa4559ca72", "native_reference_sha256": "6a130384d4ef170fe3271ba22740fa9fcf3f06937a9db63a274cd4a4d5157656", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.603146923976637, "expected_voiced_frames": 117, "missing_voiced_frames": 21, "relative_error": { "count": 96, "mean": -0.5025887999518962, "mae": 2.603146923976637, "rmse": 5.71627760441815, "abs_p95": 9.35285903619805, "abs_max": 40.1194950147727 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.1794871794871795, "expected_voiced_frames": 117, "missing_voiced_frames": 21 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.9420199394226074 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.15, "detail": { "status": "ok", "value": 0.15, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.15, "delta": 0.15 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "e97ff7851362c2039165cb37955753f1497eb355c8fa85b620deef3e8bb4094b", "native_reference_sha256": "7bff9aade99e782ef3ba9bdf6b65a01d6dabcebd2a3402c8e58868c4996d14dc", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 6.0617419089752715, "expected_voiced_frames": 117, "missing_voiced_frames": 22, "relative_error": { "count": 95, "mean": 3.6485118460596073, "mae": 6.0617419089752715, "rmse": 34.352652078093655, "abs_p95": 11.119366766337592, "abs_max": 330.8407802609064 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.18803418803418803, "expected_voiced_frames": 117, "missing_voiced_frames": 22 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.7198665142059326 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "46052d51861fb74a1be95700ad4031a521d1a9ce5fe76c268587762098b83f69", "native_reference_sha256": "6a130384d4ef170fe3271ba22740fa9fcf3f06937a9db63a274cd4a4d5157656", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.9264829263897454, "expected_voiced_frames": 117, "missing_voiced_frames": 12, "relative_error": { "count": 105, "mean": -0.9290269194372921, "mae": 2.9264829263897454, "rmse": 8.821958223675866, "abs_p95": 8.233771689901353, "abs_max": 82.40016208525881 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.10256410256410256, "expected_voiced_frames": 117, "missing_voiced_frames": 12 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.7564122676849365 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "934430303be39d2600572c3b27ec5a10c7da2e3e02492fff9eb6fb328da612ee", "native_reference_sha256": "7bff9aade99e782ef3ba9bdf6b65a01d6dabcebd2a3402c8e58868c4996d14dc", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.2652733196246095, "expected_voiced_frames": 117, "missing_voiced_frames": 11, "relative_error": { "count": 106, "mean": -0.3457280518753474, "mae": 2.2652733196246095, "rmse": 4.586192808497371, "abs_p95": 7.0849182940848525, "abs_max": 30.811666353598927 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.09401709401709402, "expected_voiced_frames": 117, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.857948064804077 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "2c0a14d656fa3567c9dc38368537769da88762f1d8bd1eb7759b0a9f41537735", "native_reference_sha256": null, "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.161055564880371 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "2a21147cd41ae8cf1681ac7d47cd6e9bae67a945b193919230afec5db21bb0b3", "native_reference_sha256": null, "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.049414873123169 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "108c92cdc660d5221b0151f0b06e787e297991a14dd9d65ee08add042d28a092", "native_reference_sha256": "2c0a14d656fa3567c9dc38368537769da88762f1d8bd1eb7759b0a9f41537735", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.455959940000507, "expected_voiced_frames": 319, "missing_voiced_frames": 41, "relative_error": { "count": 278, "mean": -0.42607946484338666, "mae": 2.455959940000507, "rmse": 4.006478833456683, "abs_p95": 10.499667163507382, "abs_max": 24.571878041513166 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12852664576802508, "expected_voiced_frames": 319, "missing_voiced_frames": 41 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4972846508026123 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "aa5c20e80aa4d9309b84ffa6e6e13f958f5e7d47d705cd85509f97ae129a6bbc", "native_reference_sha256": "2a21147cd41ae8cf1681ac7d47cd6e9bae67a945b193919230afec5db21bb0b3", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4787826502038026, "expected_voiced_frames": 319, "missing_voiced_frames": 40, "relative_error": { "count": 279, "mean": -0.5015433314534093, "mae": 2.4787826502038026, "rmse": 4.123332953418617, "abs_p95": 9.531436925159461, "abs_max": 22.33533466522802 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12539184952978055, "expected_voiced_frames": 319, "missing_voiced_frames": 40 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6654961109161377 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "bdf43e03d12027e26301fde6547acaa0e9b2f9d02b250c958506947c0a71663d", "native_reference_sha256": "2c0a14d656fa3567c9dc38368537769da88762f1d8bd1eb7759b0a9f41537735", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.1438177448749673, "expected_voiced_frames": 319, "missing_voiced_frames": 9, "relative_error": { "count": 310, "mean": -1.6726500890610578, "mae": 3.1438177448749673, "rmse": 5.299695809830368, "abs_p95": 11.61677783587684, "abs_max": 38.536831153381264 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.02821316614420063, "expected_voiced_frames": 319, "missing_voiced_frames": 9 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.9249565601348877 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "3c1d754ad393a05d23c60c508d6ac929d022537166915a94fcfa385693f4fa65", "native_reference_sha256": "2a21147cd41ae8cf1681ac7d47cd6e9bae67a945b193919230afec5db21bb0b3", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.0540918709295957, "expected_voiced_frames": 319, "missing_voiced_frames": 10, "relative_error": { "count": 309, "mean": -1.4141746010644525, "mae": 3.0540918709295957, "rmse": 4.9678176907670455, "abs_p95": 11.037195744225954, "abs_max": 28.158376679071306 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03134796238244514, "expected_voiced_frames": 319, "missing_voiced_frames": 10 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.1362924575805664 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "64142a0704ebcd7e46a990da2b7126f3d88884e793f5287548fa9fe90b6c4d45", "native_reference_sha256": null, "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7355642318725586 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "850e24c72c52e6d61b67acd28c07ccf83d6d2d38708626df49176dc638b4079e", "native_reference_sha256": null, "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.657235622406006 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "3c8cce327674d6b982384dbe1c1023abee975ad80c53e236e5c7c9673377e060", "native_reference_sha256": "64142a0704ebcd7e46a990da2b7126f3d88884e793f5287548fa9fe90b6c4d45", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.1992891404406865, "expected_voiced_frames": 101, "missing_voiced_frames": 26, "relative_error": { "count": 75, "mean": -0.45976393878431937, "mae": 2.1992891404406865, "rmse": 3.469852041058777, "abs_p95": 6.622764623890196, "abs_max": 15.164641842881029 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25742574257425743, "expected_voiced_frames": 101, "missing_voiced_frames": 26 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6110925674438477 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "457f31848de2fa59f93ec0d0a0e9af6af47a98d36342215e91dda378224f9246", "native_reference_sha256": "850e24c72c52e6d61b67acd28c07ccf83d6d2d38708626df49176dc638b4079e", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.219665073820892, "expected_voiced_frames": 103, "missing_voiced_frames": 25, "relative_error": { "count": 78, "mean": -0.41740532489166726, "mae": 2.219665073820892, "rmse": 3.4359955693121513, "abs_p95": 6.569250824918322, "abs_max": 15.363798005213118 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.24271844660194175, "expected_voiced_frames": 103, "missing_voiced_frames": 25 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6067724227905273 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "4e42dc407b50b4a18aaab6a4427849647c46b4a87f40a8d6bdb328d29722b70c", "native_reference_sha256": "64142a0704ebcd7e46a990da2b7126f3d88884e793f5287548fa9fe90b6c4d45", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.495310864822625, "expected_voiced_frames": 101, "missing_voiced_frames": 1, "relative_error": { "count": 100, "mean": -0.35578703742187234, "mae": 3.495310864822625, "rmse": 4.383752114811179, "abs_p95": 8.707415465525212, "abs_max": 14.038311265186683 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.009900990099009901, "expected_voiced_frames": 101, "missing_voiced_frames": 1 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2838759422302246 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "status": "recorded", "waveform_sha256": "3c88ce3a9611c3d258608d755cf632e48d1fa9bc24735aeee23fc6d108cedea7", "native_reference_sha256": "850e24c72c52e6d61b67acd28c07ccf83d6d2d38708626df49176dc638b4079e", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.594640072805242, "expected_voiced_frames": 103, "missing_voiced_frames": 5, "relative_error": { "count": 98, "mean": 0.0479548803185271, "mae": 3.594640072805242, "rmse": 4.549456094836024, "abs_p95": 8.57563376094834, "abs_max": 12.713155382879798 } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04854368932038835, "expected_voiced_frames": 103, "missing_voiced_frames": 5 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.0839598178863525 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } ], "paired_values": [ { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": 0.05276823043823242, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "3fb4ef0208790e1d9a08e144860ff35699c058e53a100462576d2842e37c965f", "metric_row_sha256": "d24bd4c6443d64f2e8cba119a54bf12744cca205b1885962f4c0978dc0afa077", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3087401390075684 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "66fa9fe6bb6e0880a2213e700ec59f6dd8af48283c5c350d980d948bf12644f6", "metric_row_sha256": "2768f1f8bc70d11364bf977c314bbc7c829d8ced06e4e7215b002beee0bcc70f", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.255971908569336 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": 0.028347492218017578, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "3fb4ef0208790e1d9a08e144860ff35699c058e53a100462576d2842e37c965f", "metric_row_sha256": "d24bd4c6443d64f2e8cba119a54bf12744cca205b1885962f4c0978dc0afa077", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3087401390075684 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "b44e98e3c5ac8737a3b10366cda95fcbfec3b5aa87ab5f6c7b2a6bcd99dd8553", "metric_row_sha256": "6ec8206fd104741a2d269769b5a71dbf4eae8242a4fa7bba14a21072e961b9ae", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.280392646789551 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.004485130310058594, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "3fb4ef0208790e1d9a08e144860ff35699c058e53a100462576d2842e37c965f", "metric_row_sha256": "d24bd4c6443d64f2e8cba119a54bf12744cca205b1885962f4c0978dc0afa077", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3087401390075684 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "64a0d355df20f0bc11c6826f22412a88cb5259f79a7f1b115ad836c18fe43376", "metric_row_sha256": "ed357e07ac46c307eec587fdd360799244f8753687333182190f7a70c1f43893", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.313225269317627 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": 0.28195667266845703, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "e24a1c39383ef68d78409dbc3d72621b1eca9a8b62571819aaaa9f2256748bc8", "metric_row_sha256": "a773794cd2dd57c80ae16d07eb311b41fe48278db5c5b0068052aa70ed4a91a4", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4289028644561768 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "44c63bf3bc70c9df10e573995949499e4a60667e1836e05b66a3133503156c8d", "metric_row_sha256": "c6d6beb6b0c93abe39fa03bdc17fdc0bd0b5b8c9ab2b337469972f36e445b706", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.1469461917877197 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": 0.0819098949432373, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "e24a1c39383ef68d78409dbc3d72621b1eca9a8b62571819aaaa9f2256748bc8", "metric_row_sha256": "a773794cd2dd57c80ae16d07eb311b41fe48278db5c5b0068052aa70ed4a91a4", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4289028644561768 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "f89b2bdda96a657896c8d1823138a0b4672cbe31281b7f9ad1133ba335d9922c", "metric_row_sha256": "8eeb223ff60a41828d0fdb17e0e40e8b9c5c342773dfa96d48bbe15eb6091691", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3469929695129395 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": 0.023394107818603516, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "e24a1c39383ef68d78409dbc3d72621b1eca9a8b62571819aaaa9f2256748bc8", "metric_row_sha256": "a773794cd2dd57c80ae16d07eb311b41fe48278db5c5b0068052aa70ed4a91a4", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4289028644561768 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "558cbee814fcd6343fd7e49491f5030d1d1f10fd111d35b41a9befcd06d87b10", "metric_row_sha256": "481e4c35ec1c65a94997c3c4abc4aa1537e9d33858a03fbfe7d8a0234e40d539", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4055087566375732 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.029002478545938626, "unvoiced_miss_rate": 0.9335443037974706, "automatic_mos": 0.05070853233337402, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "56c62c71aeb2bf93cb796f11c244efdc3e993d6611656898be8e52c46b544aa4", "metric_row_sha256": "a0a0815c1a6b869ed0944283d2572f6987fe6dd5ab56631be404657b8e85c3a7", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4846174172578417, "expected_voiced_frames": 158, "missing_voiced_frames": 39 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2468354430379747, "expected_voiced_frames": 158, "missing_voiced_frames": 39 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5018770694732666 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "f15ca4b5126702789e289f8d8ec6bd29f659d2cd69f9b5aafabee03a1e1e68ae", "metric_row_sha256": "234af39c908cb4487f25efed44306b8ce54973cbe8d292c23de558f589be33f2", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.5136198958037803, "expected_voiced_frames": 160, "missing_voiced_frames": 38 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2375, "expected_voiced_frames": 160, "missing_voiced_frames": 38 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4511685371398926 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.0046767654491395305, "unvoiced_miss_rate": -0.1611761930969402, "automatic_mos": 0.03850269317626953, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "56c62c71aeb2bf93cb796f11c244efdc3e993d6611656898be8e52c46b544aa4", "metric_row_sha256": "a0a0815c1a6b869ed0944283d2572f6987fe6dd5ab56631be404657b8e85c3a7", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4846174172578417, "expected_voiced_frames": 158, "missing_voiced_frames": 39 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2468354430379747, "expected_voiced_frames": 158, "missing_voiced_frames": 39 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5018770694732666 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "372b199b19f5ffcb3e3144eba9e4c1630ca8a9654c8891c42d314958a8269c12", "metric_row_sha256": "284cfabd7356760d5516e87245a5c23c030eeec77daf834bb0d5a567d3a7b5f4", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.489294182706981, "expected_voiced_frames": 161, "missing_voiced_frames": 40 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2484472049689441, "expected_voiced_frames": 161, "missing_voiced_frames": 40 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.463374376296997 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.00026785560432340816, "unvoiced_miss_rate": -0.6250976715111739, "automatic_mos": -0.03588128089904785, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "56c62c71aeb2bf93cb796f11c244efdc3e993d6611656898be8e52c46b544aa4", "metric_row_sha256": "a0a0815c1a6b869ed0944283d2572f6987fe6dd5ab56631be404657b8e85c3a7", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4846174172578417, "expected_voiced_frames": 158, "missing_voiced_frames": 39 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2468354430379747, "expected_voiced_frames": 158, "missing_voiced_frames": 39 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5018770694732666 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "36295581701905a6e57570b15f8561b1cc34087a39fd8b45a88419e98085fc96", "metric_row_sha256": "c582d1a1dc72651a9774f52b5409294040574447d69184fa616812befa3b335a", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.484885272862165, "expected_voiced_frames": 162, "missing_voiced_frames": 41 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25308641975308643, "expected_voiced_frames": 162, "missing_voiced_frames": 41 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5377583503723145 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.05560625369894101, "unvoiced_miss_rate": 0.14061207609594728, "automatic_mos": -0.006547212600708008, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "1adec25b804af2d2aa5e04663d442244cc26f6381b9c3d50a25848ef7669092c", "metric_row_sha256": "9f2b51639d51f7afc2a76b71abefd92f91ba0559e2f9791645f9ab366d799114", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.45458032270452, "expected_voiced_frames": 155, "missing_voiced_frames": 34 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.21935483870967742, "expected_voiced_frames": 155, "missing_voiced_frames": 34 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.551800012588501 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "f12964dd85e940bed291b8e5d388d7c148876a334ea7a4413b375c168d80fd43", "metric_row_sha256": "2f4b1cdf9dfcf01c5695bcd979168795b4f297a1a619a865e22419d80cc31261", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.510186576403461, "expected_voiced_frames": 156, "missing_voiced_frames": 34 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.21794871794871795, "expected_voiced_frames": 156, "missing_voiced_frames": 34 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.558347225189209 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.06013336564400884, "unvoiced_miss_rate": 0.27943291555372896, "automatic_mos": -0.08748412132263184, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "1adec25b804af2d2aa5e04663d442244cc26f6381b9c3d50a25848ef7669092c", "metric_row_sha256": "9f2b51639d51f7afc2a76b71abefd92f91ba0559e2f9791645f9ab366d799114", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.45458032270452, "expected_voiced_frames": 155, "missing_voiced_frames": 34 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.21935483870967742, "expected_voiced_frames": 155, "missing_voiced_frames": 34 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.551800012588501 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "85a3e6de38bf54142b9957e8b32eab48d645edf50c9aaf839febecc2772d153d", "metric_row_sha256": "fafde2fd081323d3b438aae74580b2b10cbcb11b147d2df3c37d95355f46e132", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.514713688348529, "expected_voiced_frames": 157, "missing_voiced_frames": 34 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.21656050955414013, "expected_voiced_frames": 157, "missing_voiced_frames": 34 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.639284133911133 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.059873740234680284, "unvoiced_miss_rate": 0.27943291555372896, "automatic_mos": -0.0210113525390625, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "1adec25b804af2d2aa5e04663d442244cc26f6381b9c3d50a25848ef7669092c", "metric_row_sha256": "9f2b51639d51f7afc2a76b71abefd92f91ba0559e2f9791645f9ab366d799114", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.45458032270452, "expected_voiced_frames": 155, "missing_voiced_frames": 34 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.21935483870967742, "expected_voiced_frames": 155, "missing_voiced_frames": 34 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.551800012588501 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "89066009f16527f1236a4f3de5d8d3968d76a6ef510e02ffbb6689d0adfaaf6d", "metric_row_sha256": "25b01c7623a4e748d5ade2885368f2344ea247b6f14cc10ec725222d2d29ab13", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.5144540629392003, "expected_voiced_frames": 157, "missing_voiced_frames": 34 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.21656050955414013, "expected_voiced_frames": 157, "missing_voiced_frames": 34 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5728113651275635 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.10406009222978474, "unvoiced_miss_rate": 0.7120253164556958, "automatic_mos": -0.06428241729736328, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "5a237dceec200dc9344c94fa85baf7d0b853f5160902a9b00634b4f1ce76f948", "metric_row_sha256": "2fcf3642aab3afc4aebf7ee2326307d90e9153162f556531fe6a76db9a71fe6a", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.80379499561333, "expected_voiced_frames": 158, "missing_voiced_frames": 11 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.06962025316455696, "expected_voiced_frames": 158, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.343353748321533 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "92bbaffce17b1bd4f39af18f13e049fab44420698441539ebe1084d24a360561", "metric_row_sha256": "0a99e7da073d78e67c47d2c874c90326412d09cbea02720943e9cb20a52780ff", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.9078550878431146, "expected_voiced_frames": 160, "missing_voiced_frames": 10 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.0625, "expected_voiced_frames": 160, "missing_voiced_frames": 10 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4076361656188965 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.028879103434380493, "unvoiced_miss_rate": 0.1297271798097327, "automatic_mos": -0.09504437446594238, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "5a237dceec200dc9344c94fa85baf7d0b853f5160902a9b00634b4f1ce76f948", "metric_row_sha256": "2fcf3642aab3afc4aebf7ee2326307d90e9153162f556531fe6a76db9a71fe6a", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.80379499561333, "expected_voiced_frames": 158, "missing_voiced_frames": 11 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.06962025316455696, "expected_voiced_frames": 158, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.343353748321533 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "c22a94e6d2092e0cdba61b2e9a2d6c67826aefaa0ab7a43735a9004dc602d10e", "metric_row_sha256": "665d4d7f3f308811161dd36a6479b2a49340cd5928f3d63611fe01184b840384", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.8326740990477104, "expected_voiced_frames": 161, "missing_voiced_frames": 11 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.06832298136645963, "expected_voiced_frames": 161, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4383981227874756 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.060319317246977366, "unvoiced_miss_rate": 0.17190185966557275, "automatic_mos": -0.2545626163482666, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "5a237dceec200dc9344c94fa85baf7d0b853f5160902a9b00634b4f1ce76f948", "metric_row_sha256": "2fcf3642aab3afc4aebf7ee2326307d90e9153162f556531fe6a76db9a71fe6a", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.80379499561333, "expected_voiced_frames": 158, "missing_voiced_frames": 11 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.06962025316455696, "expected_voiced_frames": 158, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.343353748321533 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "0010925b8b161a0ce29017235f349eabd0c3dd576a2e1a3b3d1ce38bbce9b5a8", "metric_row_sha256": "074f78a92707607c40fbe083e08ad14727a06bfdf94386dcb060a82055b9efe6", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.8641143128603073, "expected_voiced_frames": 162, "missing_voiced_frames": 11 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.06790123456790123, "expected_voiced_frames": 162, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5979163646698 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.06457640011120702, "unvoiced_miss_rate": -0.6079404466501244, "automatic_mos": 0.1180269718170166, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "60319a6cc33fa21fc4c1ce28cc426593207331e54b596f15873669c7ad36691d", "metric_row_sha256": "1a5537a3633b6ca9095f0de00e63786215a4b90b235d3ae01c7fc98253395a81", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.814166483934752, "expected_voiced_frames": 155, "missing_voiced_frames": 8 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.05161290322580645, "expected_voiced_frames": 155, "missing_voiced_frames": 8 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.217778205871582 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "dd6507d9d930d7aa7d7f1c3fde3838d460cb15162848328f52a9e2abd0f801fc", "metric_row_sha256": "27a61ba8c3f1fe948a1e0350407e94aae34ac9ca898eb6f853b52d46638f986e", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.878742884045959, "expected_voiced_frames": 156, "missing_voiced_frames": 9 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.057692307692307696, "expected_voiced_frames": 156, "missing_voiced_frames": 9 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.0997512340545654 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.07911565398044473, "unvoiced_miss_rate": -0.571193753852476, "automatic_mos": 0.016938209533691406, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "60319a6cc33fa21fc4c1ce28cc426593207331e54b596f15873669c7ad36691d", "metric_row_sha256": "1a5537a3633b6ca9095f0de00e63786215a4b90b235d3ae01c7fc98253395a81", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.814166483934752, "expected_voiced_frames": 155, "missing_voiced_frames": 8 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.05161290322580645, "expected_voiced_frames": 155, "missing_voiced_frames": 8 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.217778205871582 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "f7abbf66839bd3897b5719d9e2fd9547e51a1aa9dae96809bcc7c595f7a963ee", "metric_row_sha256": "9b06f69e349dbad0bd405fa835e0614d51c712a1fa8867bf6ba90be9d3767d8f", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.8932821379151967, "expected_voiced_frames": 157, "missing_voiced_frames": 9 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.05732484076433121, "expected_voiced_frames": 157, "missing_voiced_frames": 9 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2008399963378906 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.0558131446253447, "unvoiced_miss_rate": 0.06574892130675955, "automatic_mos": -0.07877302169799805, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "60319a6cc33fa21fc4c1ce28cc426593207331e54b596f15873669c7ad36691d", "metric_row_sha256": "1a5537a3633b6ca9095f0de00e63786215a4b90b235d3ae01c7fc98253395a81", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.814166483934752, "expected_voiced_frames": 155, "missing_voiced_frames": 8 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.05161290322580645, "expected_voiced_frames": 155, "missing_voiced_frames": 8 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.217778205871582 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "331f6cc53fe945354bc2a0e36c458c4eb6ec5e931c4b95131f4659ca52938ebf", "metric_row_sha256": "4221b6d4ef178bb7c80f9280920d2d15b6378315c1de158df73b94f83371e20b", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.8699796285600967, "expected_voiced_frames": 157, "missing_voiced_frames": 8 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.050955414012738856, "expected_voiced_frames": 157, "missing_voiced_frames": 8 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.29655122756958 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": 0.02385544776916504, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "35d484a333cf1dab94d3febcb9d0d2105a85804fe4db39537a8f57a5e5843fb3", "metric_row_sha256": "3d7ca92aed0a0c7d0c02c975d7f384bf4894d31c65c99c0f2d0095dad090a537", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.449693441390991 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "d3e29b6ecedbb29f259d2b5da2790a9fb769b78cda50fe1daae75d47da0e9713", "metric_row_sha256": "df7eaf1152f42677ba7671556d228d18e248e15f63d36cb8c50f001c7d670062", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.425837993621826 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.05512523651123047, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "35d484a333cf1dab94d3febcb9d0d2105a85804fe4db39537a8f57a5e5843fb3", "metric_row_sha256": "3d7ca92aed0a0c7d0c02c975d7f384bf4894d31c65c99c0f2d0095dad090a537", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.449693441390991 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "b5986f69364cea1af9ca0584d106968a708d0303cff0b3a963236f8a1050ddf1", "metric_row_sha256": "a87ec44484afc22d079ccb6aae5c8844162bf84f5d80f523c5c402f34a754000", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5048186779022217 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.05938243865966797, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "35d484a333cf1dab94d3febcb9d0d2105a85804fe4db39537a8f57a5e5843fb3", "metric_row_sha256": "3d7ca92aed0a0c7d0c02c975d7f384bf4894d31c65c99c0f2d0095dad090a537", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.449693441390991 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "2b1046ba58dad230b4eb3350850b571682f9f862c5cb9f88b6ea1863ade21b3b", "metric_row_sha256": "d208f3035c5eb1baf59ba6626f50c476a12d54d8926a228458843258621fab99", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.509075880050659 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": 0.16899585723876953, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "e246a24366605e5de6ce6d89e9174673cfd138c8d278c74cb046ac0049b23348", "metric_row_sha256": "83863be411ac4e175fcbedf9a0e5ca04d02e94d7c8c391432275f50abca4f5e7", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2923033237457275 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "ce73d360c6561f425030780b9bb41936ba0db8217858c516d53dc30f465c65c5", "metric_row_sha256": "dfb8a3b468a366038a97d37e7495b659c8697e09416423c2da64dc06e6fbdac2", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.123307466506958 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": 0.07678699493408203, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "e246a24366605e5de6ce6d89e9174673cfd138c8d278c74cb046ac0049b23348", "metric_row_sha256": "83863be411ac4e175fcbedf9a0e5ca04d02e94d7c8c391432275f50abca4f5e7", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2923033237457275 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "a015d2fcc1eec56bfa3722ede0fb22d357d17264b5bbdaf28a79d04b866edec0", "metric_row_sha256": "78cde8ff826ff7a14b675884b9540da5e17bbb38ee3f1793ecf8190ace277a13", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2155163288116455 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.006285429000854492, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "e246a24366605e5de6ce6d89e9174673cfd138c8d278c74cb046ac0049b23348", "metric_row_sha256": "83863be411ac4e175fcbedf9a0e5ca04d02e94d7c8c391432275f50abca4f5e7", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2923033237457275 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "69b6cc09a19df04c7917894c44a2e69314ec8cb40ea0cd5a6a6cf505992bd07c", "metric_row_sha256": "50c0bcd3057c15bf9869b1fcf9c78d307748d0ce8393290d546f9bccb6f3f312", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.298588752746582 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.042932299828181364, "unvoiced_miss_rate": 0, "automatic_mos": 0.17781567573547363, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "8056eedc68e2ad58640646f77a5c4df4ce9facdfefb2693013f1a48ef4a760bd", "metric_row_sha256": "219e6a085f79950e413b0cb0ec138c373c6174547814e3b1d120153ff3baaa5b", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.1716787977691494, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2198581560283688, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.775503158569336 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "1bc1d3429be451315e5be2fe6196e4bd6667a0145169ab797717f13c9adc67ca", "metric_row_sha256": "5f6e7ff7ec69a1b4b62d1454d0dba03019ffbfdfd12eef79aca384e89b8d5982", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.2146110975973308, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2198581560283688, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5976874828338623 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.003799839499764257, "unvoiced_miss_rate": 0, "automatic_mos": 0.06387615203857422, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "8056eedc68e2ad58640646f77a5c4df4ce9facdfefb2693013f1a48ef4a760bd", "metric_row_sha256": "219e6a085f79950e413b0cb0ec138c373c6174547814e3b1d120153ff3baaa5b", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.1716787977691494, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2198581560283688, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.775503158569336 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "feb1ee37fd381b6aba4acae0489a8c7079ba7dd637745b8c055d541bac4468e3", "metric_row_sha256": "3f66c3ad7ea0378e04e21b70aea4ae3b6f7c17046833f80515024019f6a02ac8", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.167878958269385, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2198581560283688, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7116270065307617 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 1.2982751629309064e-06, "unvoiced_miss_rate": 0, "automatic_mos": 0.02052927017211914, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "8056eedc68e2ad58640646f77a5c4df4ce9facdfefb2693013f1a48ef4a760bd", "metric_row_sha256": "219e6a085f79950e413b0cb0ec138c373c6174547814e3b1d120153ff3baaa5b", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.1716787977691494, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2198581560283688, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.775503158569336 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "e9e7cf6d978ef93e71fc7009d1de5523516949eb8174f493ee618e3d6073d6bb", "metric_row_sha256": "57e23717da0d7c7978abc33ca099bd3cbcccaca943042ad67af7727c3c065bc2", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.1716774994939865, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2198581560283688, "expected_voiced_frames": 141, "missing_voiced_frames": 31 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.754973888397217 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.023727959734211712, "unvoiced_miss_rate": -0.7516293230578952, "automatic_mos": 0.29399681091308594, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "bf380796a37f19993641182d099ce967d53b205d3d64c13344855ed440f83d0a", "metric_row_sha256": "5512a0b995c5d57cd1e66e7baa384811dca88423f2a58488dcf4ebc8963c1a2f", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.423271972110417, "expected_voiced_frames": 143, "missing_voiced_frames": 32 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.22377622377622378, "expected_voiced_frames": 143, "missing_voiced_frames": 32 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.8541226387023926 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "4f69304610d511311c2936f8d8d81b7737e0de2128026f1656e4cc4a57a02522", "metric_row_sha256": "881c099975424b2359a4f66dd4d5981af0a7eab747a9c1a7d8a6a5adb4876118", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.399544012376205, "expected_voiced_frames": 147, "missing_voiced_frames": 34 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.23129251700680273, "expected_voiced_frames": 147, "missing_voiced_frames": 34 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5601258277893066 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.008717161554346653, "unvoiced_miss_rate": 0, "automatic_mos": 0.06101727485656738, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "bf380796a37f19993641182d099ce967d53b205d3d64c13344855ed440f83d0a", "metric_row_sha256": "5512a0b995c5d57cd1e66e7baa384811dca88423f2a58488dcf4ebc8963c1a2f", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.423271972110417, "expected_voiced_frames": 143, "missing_voiced_frames": 32 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.22377622377622378, "expected_voiced_frames": 143, "missing_voiced_frames": 32 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.8541226387023926 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "d91225ad0a3d20e676e120b58fc3bfa251c488d95fc031dd52b70cd2682275ce", "metric_row_sha256": "08c30fdf1c58796b3698454af4d617bf82e69d8ccb07a7b72368d482f018f3c5", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.41455481055607, "expected_voiced_frames": 143, "missing_voiced_frames": 32 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.22377622377622378, "expected_voiced_frames": 143, "missing_voiced_frames": 32 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.793105363845825 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -4.143579599436009e-07, "unvoiced_miss_rate": 0, "automatic_mos": 0.015845775604248047, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "bf380796a37f19993641182d099ce967d53b205d3d64c13344855ed440f83d0a", "metric_row_sha256": "5512a0b995c5d57cd1e66e7baa384811dca88423f2a58488dcf4ebc8963c1a2f", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.423271972110417, "expected_voiced_frames": 143, "missing_voiced_frames": 32 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.22377622377622378, "expected_voiced_frames": 143, "missing_voiced_frames": 32 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.8541226387023926 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "44010ecd3e30fea37fc314d75707781e1394fc92a8acbdf76317d19ddbb27257", "metric_row_sha256": "a5ee77a0ffddce82b38982677cefb11c2dca7820748b1c41df5aa86482da8605", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4232723864683767, "expected_voiced_frames": 143, "missing_voiced_frames": 32 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.22377622377622378, "expected_voiced_frames": 143, "missing_voiced_frames": 32 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.8382768630981445 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.10683200077045196, "unvoiced_miss_rate": -0.7092198581560281, "automatic_mos": 0.2629280090332031, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "69a04884103d8af8de3dbe10aac65d1c74e7af005e709d5c438ed16a16010549", "metric_row_sha256": "712c9d66b91550802dace2dd01be6281391a50691a06bc198ce7659442bf494a", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.052278100448317, "expected_voiced_frames": 141, "missing_voiced_frames": 5 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03546099290780142, "expected_voiced_frames": 141, "missing_voiced_frames": 5 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.459120273590088 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "65085023011d3842b8018ac786d6928eb5668e00256ac74ce689082a6751bae7", "metric_row_sha256": "375297503f1db66bf0807b1d39912cac9e6c67f0c64465058abdd498d6eda6a3", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.159110101218769, "expected_voiced_frames": 141, "missing_voiced_frames": 6 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.0425531914893617, "expected_voiced_frames": 141, "missing_voiced_frames": 6 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.1961922645568848 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.012011203997654096, "unvoiced_miss_rate": 0, "automatic_mos": 0.2291431427001953, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "69a04884103d8af8de3dbe10aac65d1c74e7af005e709d5c438ed16a16010549", "metric_row_sha256": "712c9d66b91550802dace2dd01be6281391a50691a06bc198ce7659442bf494a", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.052278100448317, "expected_voiced_frames": 141, "missing_voiced_frames": 5 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03546099290780142, "expected_voiced_frames": 141, "missing_voiced_frames": 5 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.459120273590088 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "7f6495887d5f5d47e2cbd5728681bcf0343669b8ba4f5f086823412ee6e031c8", "metric_row_sha256": "2a503f25d5c5818e0e47d13871a523b6d3cba17b915bd80e200f160f060e7cb8", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.040266896450663, "expected_voiced_frames": 141, "missing_voiced_frames": 5 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03546099290780142, "expected_voiced_frames": 141, "missing_voiced_frames": 5 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2299771308898926 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.00016233919641006622, "unvoiced_miss_rate": 0, "automatic_mos": 0.12546372413635254, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "69a04884103d8af8de3dbe10aac65d1c74e7af005e709d5c438ed16a16010549", "metric_row_sha256": "712c9d66b91550802dace2dd01be6281391a50691a06bc198ce7659442bf494a", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.052278100448317, "expected_voiced_frames": 141, "missing_voiced_frames": 5 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03546099290780142, "expected_voiced_frames": 141, "missing_voiced_frames": 5 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.459120273590088 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "c6c06a298030eca3f7adb7fdf9851e5323bd99bf8813b2cc79fccfaaa810569f", "metric_row_sha256": "9ecaa14f6ac9b885aa53f748b5edfd3521094d214c83d7b2eb2e1bb280e09795", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.052115761251907, "expected_voiced_frames": 141, "missing_voiced_frames": 5 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03546099290780142, "expected_voiced_frames": 141, "missing_voiced_frames": 5 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3336565494537354 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.28045688860344287, "unvoiced_miss_rate": -3.2681604110175533, "automatic_mos": 0.25576114654541016, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "c1aa72cdd4b91ad41084a4b253e11bd6f9fc9d86f47d044b9609943ff90929a0", "metric_row_sha256": "95953f2a4c612fe30ef4c9bb71f7cb9f5c8e43f63810d7ea3cd83612c90c8221", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.27440482768568, "expected_voiced_frames": 143, "missing_voiced_frames": 7 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04895104895104895, "expected_voiced_frames": 143, "missing_voiced_frames": 7 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4007201194763184 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "09f09ae312805ea0f54ca68227aea0ca8c306e6f44376919ecd16456b1289267", "metric_row_sha256": "e733b0144d3989a9238c4165a992fbc4cf41c1022553fd24b6644ca4965f6295", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.993947939082237, "expected_voiced_frames": 147, "missing_voiced_frames": 12 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.08163265306122448, "expected_voiced_frames": 147, "missing_voiced_frames": 12 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.144958972930908 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.17884948626291663, "unvoiced_miss_rate": -1.3986013986013992, "automatic_mos": 0.21976709365844727, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "c1aa72cdd4b91ad41084a4b253e11bd6f9fc9d86f47d044b9609943ff90929a0", "metric_row_sha256": "95953f2a4c612fe30ef4c9bb71f7cb9f5c8e43f63810d7ea3cd83612c90c8221", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.27440482768568, "expected_voiced_frames": 143, "missing_voiced_frames": 7 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04895104895104895, "expected_voiced_frames": 143, "missing_voiced_frames": 7 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4007201194763184 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "653056c6e110bfae7b79eafd8b68096b60900b42b7ea51adf633d36242001a34", "metric_row_sha256": "0d6461dd1bb923837ffee33e17cd7e38c805880b1163830699a355167cb4ba42", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.0955553414227635, "expected_voiced_frames": 143, "missing_voiced_frames": 9 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.06293706293706294, "expected_voiced_frames": 143, "missing_voiced_frames": 9 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.180953025817871 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.00013476829886460706, "unvoiced_miss_rate": 0, "automatic_mos": 0.2528233528137207, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "c1aa72cdd4b91ad41084a4b253e11bd6f9fc9d86f47d044b9609943ff90929a0", "metric_row_sha256": "95953f2a4c612fe30ef4c9bb71f7cb9f5c8e43f63810d7ea3cd83612c90c8221", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.27440482768568, "expected_voiced_frames": 143, "missing_voiced_frames": 7 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04895104895104895, "expected_voiced_frames": 143, "missing_voiced_frames": 7 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4007201194763184 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "9a89cff025a29917a37ef97f4c3544043d2d5ecb3070ecaceadfeac91bf9435d", "metric_row_sha256": "09ee7bf57d8edafe3f74c74e09f01cbfb88a894d58d788481634a51fe3cf9a27", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.2745395959845447, "expected_voiced_frames": 143, "missing_voiced_frames": 7 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04895104895104895, "expected_voiced_frames": 143, "missing_voiced_frames": 7 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.1478967666625977 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.06357836723327637, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "6a130384d4ef170fe3271ba22740fa9fcf3f06937a9db63a274cd4a4d5157656", "metric_row_sha256": "ff8994b2f54f23e79ef681ea7ad0b8f8c52c70bbce801a9d46b8e32fd006d5a9", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2409791946411133 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "bee6b53a7fe774160c03a93083e906a9b38cf640e006db02ed7ff8660b282ace", "metric_row_sha256": "f49a8867dd2884693683d1bcf56ddd7ebe3117a40ebc7842eaeb5e70bab09fef", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3045575618743896 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.1296677589416504, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "6a130384d4ef170fe3271ba22740fa9fcf3f06937a9db63a274cd4a4d5157656", "metric_row_sha256": "ff8994b2f54f23e79ef681ea7ad0b8f8c52c70bbce801a9d46b8e32fd006d5a9", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2409791946411133 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "b39de8c9555ba9c5ece7bf31410698367738566f4162ba15b4a1b1ff746d4c62", "metric_row_sha256": "00915e297e444eeda23645c23a479015f26e441459f7caaf0f55208e12c74a48", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3706469535827637 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.15574860572814941, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "6a130384d4ef170fe3271ba22740fa9fcf3f06937a9db63a274cd4a4d5157656", "metric_row_sha256": "ff8994b2f54f23e79ef681ea7ad0b8f8c52c70bbce801a9d46b8e32fd006d5a9", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2409791946411133 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "6cf3876756c3750b646ec255f2594c3fa5e5030e632c9bbeb1c8b015d9d660d4", "metric_row_sha256": "1fcc9bc5ab9a61a0ea6e2f9e8a76a414593f97c80da2697f5ed1be3e4d516a85", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3967278003692627 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.2720189094543457, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": true, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags_applied_only_to_PML": true, "harm_observed": true, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "7bff9aade99e782ef3ba9bdf6b65a01d6dabcebd2a3402c8e58868c4996d14dc", "metric_row_sha256": "e1746738fe93de9c61313a7a941470f4e16378b66a6d18e3eae217e48772c26c", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.038602828979492 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "3c80cb0183665bde7489d084712b09a3f802d74c5f31cb9ac1ea25c0dd609bb9", "metric_row_sha256": "f9f67739fb9d8f59b0ac4efaece9db5e632ba60d79a5f98017305a07ac0fb852", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.310621738433838 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.11609315872192383, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "7bff9aade99e782ef3ba9bdf6b65a01d6dabcebd2a3402c8e58868c4996d14dc", "metric_row_sha256": "e1746738fe93de9c61313a7a941470f4e16378b66a6d18e3eae217e48772c26c", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.038602828979492 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "9e304a730282597644a2cb98f8463de959fe67e82ec538ac95b24713eadaae41", "metric_row_sha256": "41f9759dba17ffe58818e6a9faf30e85982e4e37d69c2b10c021072bcf6fe963", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.154695987701416 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.15077733993530273, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "7bff9aade99e782ef3ba9bdf6b65a01d6dabcebd2a3402c8e58868c4996d14dc", "metric_row_sha256": "e1746738fe93de9c61313a7a941470f4e16378b66a6d18e3eae217e48772c26c", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.038602828979492 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "23f5ecac6f0600c8b612ec2e5464efb96be2ced7d9e8d316af0ada76e8e46f93", "metric_row_sha256": "8883786c03a45c9ff29d46e280bee064becae8e3fd6ae2fa0687d39c0eb5ccfc", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.189380168914795 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.0570778258773208, "unvoiced_miss_rate": -0.8547008547008544, "automatic_mos": 0.08728814125061035, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.15, "delta": 0.15 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.15, "delta": 0.15 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "6d4606089c4c9428cebe258d7036c74c04e13b155312e7bc4e5c03fa4559ca72", "metric_row_sha256": "5d6bdc233952114a167ae060712347c20621c8eda1e13f2e690560422014c506", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.603146923976637, "expected_voiced_frames": 117, "missing_voiced_frames": 21 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.1794871794871795, "expected_voiced_frames": 117, "missing_voiced_frames": 21 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.9420199394226074 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.15, "detail": { "status": "ok", "value": 0.15, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.15, "delta": 0.15 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "b5d58d1a3452c7cced18f708f3b23c109acb016e90f0ab159606d447fa2a8ba2", "metric_row_sha256": "b6764c815672284848dfb610efcf5215f5ddd712b0ef14ea8cbee097b67280a3", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.660224749853958, "expected_voiced_frames": 117, "missing_voiced_frames": 22 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.18803418803418803, "expected_voiced_frames": 117, "missing_voiced_frames": 22 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.854731798171997 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.15, "detail": { "status": "ok", "value": 0.15, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.15, "delta": 0.15 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.04508032445365817, "unvoiced_miss_rate": -1.7094017094017089, "automatic_mos": -0.009618759155273438, "recognition": 0.04999999999999999 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.15, "delta": 0.15 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 2, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.1, "delta": 0.1 }, "candidate_minus_reference_output_rate": 0.04999999999999999 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "6d4606089c4c9428cebe258d7036c74c04e13b155312e7bc4e5c03fa4559ca72", "metric_row_sha256": "5d6bdc233952114a167ae060712347c20621c8eda1e13f2e690560422014c506", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.603146923976637, "expected_voiced_frames": 117, "missing_voiced_frames": 21 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.1794871794871795, "expected_voiced_frames": 117, "missing_voiced_frames": 21 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.9420199394226074 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.15, "detail": { "status": "ok", "value": 0.15, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.15, "delta": 0.15 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "758b0ed995d72ee4715440ce181c32f90ce018986d8010890368718a09bae037", "metric_row_sha256": "35403441b7d5823fe95683fa2c2a18327a71df954e09194ff9a56df8d33c7c3b", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.6482272484302953, "expected_voiced_frames": 117, "missing_voiced_frames": 23 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.19658119658119658, "expected_voiced_frames": 117, "missing_voiced_frames": 23 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.951638698577881 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.1, "detail": { "status": "ok", "value": 0.1, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 2, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.1, "delta": 0.1 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.03111834048292339, "unvoiced_miss_rate": -1.7094017094017089, "automatic_mos": -0.06131458282470703, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.15, "delta": 0.15 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.15, "delta": 0.15 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "6d4606089c4c9428cebe258d7036c74c04e13b155312e7bc4e5c03fa4559ca72", "metric_row_sha256": "5d6bdc233952114a167ae060712347c20621c8eda1e13f2e690560422014c506", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.603146923976637, "expected_voiced_frames": 117, "missing_voiced_frames": 21 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.1794871794871795, "expected_voiced_frames": 117, "missing_voiced_frames": 21 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.9420199394226074 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.15, "detail": { "status": "ok", "value": 0.15, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.15, "delta": 0.15 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "c8da7838eee5e846ea1ea3e638a4f7270f7be433fc9c893642421fd34bbe512e", "metric_row_sha256": "281e2df940e9c2d9acffff3a700de51ea39a2544f96458c644c285d018090901", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.6342652644595606, "expected_voiced_frames": 117, "missing_voiced_frames": 23 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.19658119658119658, "expected_voiced_frames": 117, "missing_voiced_frames": 23 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.0033345222473145 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.15, "detail": { "status": "ok", "value": 0.15, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.15, "delta": 0.15 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 3.599069484496858, "unvoiced_miss_rate": -0.327015979189893, "automatic_mos": 0.14874982833862305, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "e97ff7851362c2039165cb37955753f1497eb355c8fa85b620deef3e8bb4094b", "metric_row_sha256": "f4d8718920015f64631f0d8a033fe7f61f5008e4ca98f4750f2e0b1d6e57522b", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 6.0617419089752715, "expected_voiced_frames": 117, "missing_voiced_frames": 22 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.18803418803418803, "expected_voiced_frames": 117, "missing_voiced_frames": 22 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.7198665142059326 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "2ed85deb832fe249f779569e7c6379b57d3db0525df04a26444f4d5151db0b89", "metric_row_sha256": "b48c6181b87d007016ec63cced5ec1b1dd5921a0f288a7383e755953569e7193", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4626724244784137, "expected_voiced_frames": 115, "missing_voiced_frames": 22 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.19130434782608696, "expected_voiced_frames": 115, "missing_voiced_frames": 22 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.5711166858673096 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -1.2896231894123744, "unvoiced_miss_rate": -0.8547008547008544, "automatic_mos": -0.03597712516784668, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "e97ff7851362c2039165cb37955753f1497eb355c8fa85b620deef3e8bb4094b", "metric_row_sha256": "f4d8718920015f64631f0d8a033fe7f61f5008e4ca98f4750f2e0b1d6e57522b", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 6.0617419089752715, "expected_voiced_frames": 117, "missing_voiced_frames": 22 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.18803418803418803, "expected_voiced_frames": 117, "missing_voiced_frames": 22 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.7198665142059326 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "efd7cd6a2dca10baebb8f1cc44f3bf043cb8e0ead5c5927beb0d0ccf208f03f7", "metric_row_sha256": "911cef3626443d11ca85a08a74961896dbf05acb49284f4cea7cfa713bc9b665", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 7.351365098387646, "expected_voiced_frames": 117, "missing_voiced_frames": 23 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.19658119658119658, "expected_voiced_frames": 117, "missing_voiced_frames": 23 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.7558436393737793 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.9572004463599093, "unvoiced_miss_rate": 0, "automatic_mos": -0.04725360870361328, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "e97ff7851362c2039165cb37955753f1497eb355c8fa85b620deef3e8bb4094b", "metric_row_sha256": "f4d8718920015f64631f0d8a033fe7f61f5008e4ca98f4750f2e0b1d6e57522b", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 6.0617419089752715, "expected_voiced_frames": 117, "missing_voiced_frames": 22 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.18803418803418803, "expected_voiced_frames": 117, "missing_voiced_frames": 22 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.7198665142059326 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "5301869c730597ada086b2a0b7274654e22a93872364e3a7868e1d74098902c2", "metric_row_sha256": "d8ecd3132960a2da8a7d28978996eb72ba54df091e619fea2dd845af53788633", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 7.018942355335181, "expected_voiced_frames": 117, "missing_voiced_frames": 22 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.18803418803418803, "expected_voiced_frames": 117, "missing_voiced_frames": 22 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.767120122909546 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.47189489327017187, "unvoiced_miss_rate": 1.7094017094017089, "automatic_mos": -0.22395920753479004, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": true, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": true, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "46052d51861fb74a1be95700ad4031a521d1a9ce5fe76c268587762098b83f69", "metric_row_sha256": "13c88e4165cb24a951ffbdbdc2d679a32a0602bcef592b3c4c18984856ce0804", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.9264829263897454, "expected_voiced_frames": 117, "missing_voiced_frames": 12 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.10256410256410256, "expected_voiced_frames": 117, "missing_voiced_frames": 12 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.7564122676849365 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "082bd02b2d99751a844f503d91a3c6df5c7f98d60aef01c2c8d9ee481295c776", "metric_row_sha256": "bc59ef5e94c72f56447a4eba582a443615d71e2d65a6859b2e5c3fcf5c13e46c", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4545880331195735, "expected_voiced_frames": 117, "missing_voiced_frames": 10 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.08547008547008547, "expected_voiced_frames": 117, "missing_voiced_frames": 10 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.9803714752197266 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.3643650145073831, "unvoiced_miss_rate": 2.5641025641025634, "automatic_mos": -0.1640474796295166, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "46052d51861fb74a1be95700ad4031a521d1a9ce5fe76c268587762098b83f69", "metric_row_sha256": "13c88e4165cb24a951ffbdbdc2d679a32a0602bcef592b3c4c18984856ce0804", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.9264829263897454, "expected_voiced_frames": 117, "missing_voiced_frames": 12 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.10256410256410256, "expected_voiced_frames": 117, "missing_voiced_frames": 12 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.7564122676849365 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "1ee658832e1ddbc0fead3358ce91f9abb4c6243747772c1b1995ed0390499693", "metric_row_sha256": "6a3951d87304fe461c1dc9a901a0b3404f9de15e51fcff336bddd7264af2ddf0", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.2908479408971285, "expected_voiced_frames": 117, "missing_voiced_frames": 9 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.07692307692307693, "expected_voiced_frames": 117, "missing_voiced_frames": 9 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.920459747314453 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.20830466677898274, "unvoiced_miss_rate": 1.7094017094017089, "automatic_mos": -0.24060535430908203, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "46052d51861fb74a1be95700ad4031a521d1a9ce5fe76c268587762098b83f69", "metric_row_sha256": "13c88e4165cb24a951ffbdbdc2d679a32a0602bcef592b3c4c18984856ce0804", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.9264829263897454, "expected_voiced_frames": 117, "missing_voiced_frames": 12 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.10256410256410256, "expected_voiced_frames": 117, "missing_voiced_frames": 12 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.7564122676849365 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "448a41231e539e37d40953699caf50f78ed0866197fa9c0aa01c689ca773f9f8", "metric_row_sha256": "16ab904b2bc1488b239a7d3dda5269b9a665a500bdb3f39a768cb1de50bcf51f", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.134787593168728, "expected_voiced_frames": 117, "missing_voiced_frames": 10 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.08547008547008547, "expected_voiced_frames": 117, "missing_voiced_frames": 10 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.9970176219940186 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.22981529258747324, "unvoiced_miss_rate": 1.575622445187662, "automatic_mos": -0.20714569091796875, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": true, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": true, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "934430303be39d2600572c3b27ec5a10c7da2e3e02492fff9eb6fb328da612ee", "metric_row_sha256": "ce6dea0ab48d31a5021421b326cead1112b1122b3642ddf491a45b69c7a98e56", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.2652733196246095, "expected_voiced_frames": 117, "missing_voiced_frames": 11 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.09401709401709402, "expected_voiced_frames": 117, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.857948064804077 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "d56d4907bb8ee291d7ca7ff2c240ef4a1bdb18d485c78449b45dd152818c0e26", "metric_row_sha256": "fea4fc9d629d8103302210bba0171f5825db3209d4c4e5cdef4ec1ee26b591a0", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.495088612212083, "expected_voiced_frames": 115, "missing_voiced_frames": 9 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.0782608695652174, "expected_voiced_frames": 115, "missing_voiced_frames": 9 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.065093755722046 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -4.873747182363205, "unvoiced_miss_rate": 0.8547008547008544, "automatic_mos": -0.18715810775756836, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "934430303be39d2600572c3b27ec5a10c7da2e3e02492fff9eb6fb328da612ee", "metric_row_sha256": "ce6dea0ab48d31a5021421b326cead1112b1122b3642ddf491a45b69c7a98e56", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.2652733196246095, "expected_voiced_frames": 117, "missing_voiced_frames": 11 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.09401709401709402, "expected_voiced_frames": 117, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.857948064804077 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "4f8dc9a31dba23758b8ba40719eee80f25a3ac6ad731fa319bf29302288eac6c", "metric_row_sha256": "2a8a21c42a6b1df4d10dbdec5bd4b0428a8fd5653f2569e2086ff6f76ea47bfd", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 7.139020501987815, "expected_voiced_frames": 117, "missing_voiced_frames": 10 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.08547008547008547, "expected_voiced_frames": 117, "missing_voiced_frames": 10 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.0451061725616455 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -4.459810555870776, "unvoiced_miss_rate": 0, "automatic_mos": -0.18902277946472168, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "934430303be39d2600572c3b27ec5a10c7da2e3e02492fff9eb6fb328da612ee", "metric_row_sha256": "ce6dea0ab48d31a5021421b326cead1112b1122b3642ddf491a45b69c7a98e56", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.2652733196246095, "expected_voiced_frames": 117, "missing_voiced_frames": 11 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.09401709401709402, "expected_voiced_frames": 117, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.857948064804077 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "53a9de0adae6d1f1b59307c0e9d5345f9cbdd7fc155be3ba20e9327d5ae7085b", "metric_row_sha256": "993cd953a30ef7728e3f185f2d3e0f4010d9669ad356010b4c99550c0b56d4f5", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 6.725083875495386, "expected_voiced_frames": 117, "missing_voiced_frames": 11 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.09401709401709402, "expected_voiced_frames": 117, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.046970844268799 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.356550931930542, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": true, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags_applied_only_to_PML": true, "harm_observed": true, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "2c0a14d656fa3567c9dc38368537769da88762f1d8bd1eb7759b0a9f41537735", "metric_row_sha256": "780cbfc94d99d836504e17e1bd2f7e5d3178d3427dd6d8dd0f76a12d666d42e7", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.161055564880371 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "82cfae761e5c0f68551f01def9e993c41545f7fc8d731b60e1305811d7f63067", "metric_row_sha256": "a56d20523d715be285e487103b768b63d3cc1e1ff76b7787d33c5b967cb7a4d8", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.517606496810913 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.39362287521362305, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "2c0a14d656fa3567c9dc38368537769da88762f1d8bd1eb7759b0a9f41537735", "metric_row_sha256": "780cbfc94d99d836504e17e1bd2f7e5d3178d3427dd6d8dd0f76a12d666d42e7", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.161055564880371 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "546be5acbc36a76c03db15837211519607b7c74777715eba9bb6794184b556ca", "metric_row_sha256": "c98e3fdf2cbdaaa3c95f209d1be479bb0c2bb1993170128bb4e51d62975cd0f2", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.554678440093994 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.4857335090637207, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "2c0a14d656fa3567c9dc38368537769da88762f1d8bd1eb7759b0a9f41537735", "metric_row_sha256": "780cbfc94d99d836504e17e1bd2f7e5d3178d3427dd6d8dd0f76a12d666d42e7", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.161055564880371 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "b45a6380b5c8f17818af0dfae734ac4982a0ee6a4490f8dfa38759fe2f4e2cb5", "metric_row_sha256": "7773c9e8cbebacf4cad2ffde9d200b4e66a53ef28563af2b098491a9949789dd", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.646789073944092 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.2821645736694336, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": true, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags_applied_only_to_PML": true, "harm_observed": true, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "2a21147cd41ae8cf1681ac7d47cd6e9bae67a945b193919230afec5db21bb0b3", "metric_row_sha256": "db39bde0b5a69470f99da8ef30bcc8b7a078a65a51b931a4a636094b2db80727", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.049414873123169 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "dd0c86f3d424bc48154feff1e81ed6de6d4108469e419c8fcb13f5642277f77c", "metric_row_sha256": "473a9dac8f92425ea5456f192e6cfb229ac211e2ebe075d4e37db3bca3b0c0c4", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3315794467926025 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.3048090934753418, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "2a21147cd41ae8cf1681ac7d47cd6e9bae67a945b193919230afec5db21bb0b3", "metric_row_sha256": "db39bde0b5a69470f99da8ef30bcc8b7a078a65a51b931a4a636094b2db80727", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.049414873123169 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "e0752efd9d526392bf75afa41d5a3032ad67731c86a028a46fa73f2566f68aa2", "metric_row_sha256": "3318fd9dcd34b0a5710592c4be62be085e9c14b1b0e651c600625ef82d1fd5de", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3542239665985107 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.23065972328186035, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "2a21147cd41ae8cf1681ac7d47cd6e9bae67a945b193919230afec5db21bb0b3", "metric_row_sha256": "db39bde0b5a69470f99da8ef30bcc8b7a078a65a51b931a4a636094b2db80727", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.049414873123169 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "4b45a182c576a67cdcab118faccba5276fdf67db1ce2539efa7050809fd74bce", "metric_row_sha256": "98dd022df5af45c80e4fea3e7f04b14795e59bb10b2d67b69bc9aa479efac0c7", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2800745964050293 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.07107830130830273, "unvoiced_miss_rate": -0.1504313984296901, "automatic_mos": -0.33791208267211914, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": true, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": true, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "108c92cdc660d5221b0151f0b06e787e297991a14dd9d65ee08add042d28a092", "metric_row_sha256": "dff8113a83fb936b046493afdbac5d81a511cea275b4611516db1d05cc07f729", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.455959940000507, "expected_voiced_frames": 319, "missing_voiced_frames": 41 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12852664576802508, "expected_voiced_frames": 319, "missing_voiced_frames": 41 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4972846508026123 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "8fcc061d8e5ae305f5549bbc754d3df844c665be5f6e92cda03ad02b1c838b63", "metric_row_sha256": "0d6465bf19918572e3d8e61e7213d971760ccb1f1e556cac0098d7464cba4de1", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.52703824130881, "expected_voiced_frames": 323, "missing_voiced_frames": 42 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.13003095975232198, "expected_voiced_frames": 323, "missing_voiced_frames": 42 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.8351967334747314 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.16553574962001028, "unvoiced_miss_rate": 0.43030432835530347, "automatic_mos": -0.3081226348876953, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "108c92cdc660d5221b0151f0b06e787e297991a14dd9d65ee08add042d28a092", "metric_row_sha256": "dff8113a83fb936b046493afdbac5d81a511cea275b4611516db1d05cc07f729", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.455959940000507, "expected_voiced_frames": 319, "missing_voiced_frames": 41 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12852664576802508, "expected_voiced_frames": 319, "missing_voiced_frames": 41 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4972846508026123 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "4f9b1ae812443c7b4f5996d20cf2dca1b5622345b8ec3600e104661997ce80ab", "metric_row_sha256": "3a29c27574744087fbfa8ecb11707d1bb239a3d86af8a6dcda416fd82edced16", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.6214956896205175, "expected_voiced_frames": 322, "missing_voiced_frames": 40 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12422360248447205, "expected_voiced_frames": 322, "missing_voiced_frames": 40 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.8054072856903076 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.12804266248917884, "unvoiced_miss_rate": 0.11974532214412414, "automatic_mos": -0.28577470779418945, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "108c92cdc660d5221b0151f0b06e787e297991a14dd9d65ee08add042d28a092", "metric_row_sha256": "dff8113a83fb936b046493afdbac5d81a511cea275b4611516db1d05cc07f729", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.455959940000507, "expected_voiced_frames": 319, "missing_voiced_frames": 41 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12852664576802508, "expected_voiced_frames": 319, "missing_voiced_frames": 41 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4972846508026123 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "28d85162fd6360c67380745db8d1e999c220ab137c620b9e38a89f9deab98eff", "metric_row_sha256": "00bd830c69daf9e6ad871b0c784586fb5e661e8e74e5ce7ff3a9d966648fc63e", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.584002602489686, "expected_voiced_frames": 322, "missing_voiced_frames": 41 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12732919254658384, "expected_voiced_frames": 322, "missing_voiced_frames": 41 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7830593585968018 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.1824525159193735, "unvoiced_miss_rate": -0.15431349903432812, "automatic_mos": -0.16655492782592773, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": true, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": true, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "aa5c20e80aa4d9309b84ffa6e6e13f958f5e7d47d705cd85509f97ae129a6bbc", "metric_row_sha256": "7beb8e0b33aa8f9a0f676758e0a4c9c8e40624c05e81cdd40f96e0e30318390c", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4787826502038026, "expected_voiced_frames": 319, "missing_voiced_frames": 40 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12539184952978055, "expected_voiced_frames": 319, "missing_voiced_frames": 40 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6654961109161377 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "b28b91f45e924e38cd9dab10984e28bb2421ad7945e0fefeddb5e1730d69ad26", "metric_row_sha256": "7295f0b84cef807d039c4a2f4580a7b744b141223c292071cffc577e80c56a0b", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.661235166123176, "expected_voiced_frames": 323, "missing_voiced_frames": 41 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12693498452012383, "expected_voiced_frames": 323, "missing_voiced_frames": 41 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.8320510387420654 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.14696900033320048, "unvoiced_miss_rate": 0.7011787224484602, "automatic_mos": -0.18094182014465332, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "aa5c20e80aa4d9309b84ffa6e6e13f958f5e7d47d705cd85509f97ae129a6bbc", "metric_row_sha256": "7beb8e0b33aa8f9a0f676758e0a4c9c8e40624c05e81cdd40f96e0e30318390c", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4787826502038026, "expected_voiced_frames": 319, "missing_voiced_frames": 40 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12539184952978055, "expected_voiced_frames": 319, "missing_voiced_frames": 40 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6654961109161377 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "9c55bd5513136d74c0632debc3da1389eaab09c5b8c6e2eeb967247bc652979b", "metric_row_sha256": "7e57b8682d9d8fc295dcea06e834c31cc24ed8ada9ef300d4e4f17e8533f1d2c", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.625751650537003, "expected_voiced_frames": 321, "missing_voiced_frames": 38 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.11838006230529595, "expected_voiced_frames": 321, "missing_voiced_frames": 38 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.846437931060791 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.13952992159947541, "unvoiced_miss_rate": 0.11682470453085042, "automatic_mos": -0.23460102081298828, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "aa5c20e80aa4d9309b84ffa6e6e13f958f5e7d47d705cd85509f97ae129a6bbc", "metric_row_sha256": "7beb8e0b33aa8f9a0f676758e0a4c9c8e40624c05e81cdd40f96e0e30318390c", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4787826502038026, "expected_voiced_frames": 319, "missing_voiced_frames": 40 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12539184952978055, "expected_voiced_frames": 319, "missing_voiced_frames": 40 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6654961109161377 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "7c14268757b7b8b63cbd5e3a648036dadd57a8ed84b404b863dae6fb37a75e93", "metric_row_sha256": "aec742504fd7994bf5e5c5b6b039dfae45d1d1968dc46b45d9390a21759214a7", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.618312571803278, "expected_voiced_frames": 322, "missing_voiced_frames": 40 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12422360248447205, "expected_voiced_frames": 322, "missing_voiced_frames": 40 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.900097131729126 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.03509316792477035, "unvoiced_miss_rate": -0.893853664217708, "automatic_mos": -0.4072532653808594, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": true, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": true, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "bdf43e03d12027e26301fde6547acaa0e9b2f9d02b250c958506947c0a71663d", "metric_row_sha256": "76d9713274034bd613d5513c2fb85ac96a1a49be9591ce61f3989e60ae1afde2", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.1438177448749673, "expected_voiced_frames": 319, "missing_voiced_frames": 9 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.02821316614420063, "expected_voiced_frames": 319, "missing_voiced_frames": 9 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.9249565601348877 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "831611555a330d1412fc5497d46f47073c8633eb40f0513b842549b680771bea", "metric_row_sha256": "8f5367a9b6e74dc5c58ce411018f61cf4869219b9035cd277cfacd500a5af958", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.108724576950197, "expected_voiced_frames": 323, "missing_voiced_frames": 12 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03715170278637771, "expected_voiced_frames": 323, "missing_voiced_frames": 12 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.332209825515747 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.005685113602457825, "unvoiced_miss_rate": -0.28427344769173835, "automatic_mos": -0.5261802673339844, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "bdf43e03d12027e26301fde6547acaa0e9b2f9d02b250c958506947c0a71663d", "metric_row_sha256": "76d9713274034bd613d5513c2fb85ac96a1a49be9591ce61f3989e60ae1afde2", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.1438177448749673, "expected_voiced_frames": 319, "missing_voiced_frames": 9 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.02821316614420063, "expected_voiced_frames": 319, "missing_voiced_frames": 9 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.9249565601348877 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "3ab09914cd862b8bc66ba18d8d21ecf9a03b6609fa2dd7e5a50e5a839629aea3", "metric_row_sha256": "74c8d61c23e0eba91f523a95ea21e8dcc0d0276b1efea3c7262e6074dd91a31d", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.1381326312725095, "expected_voiced_frames": 322, "missing_voiced_frames": 10 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.031055900621118012, "expected_voiced_frames": 322, "missing_voiced_frames": 10 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.451136827468872 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.011658066209697893, "unvoiced_miss_rate": -0.5948324539029187, "automatic_mos": -0.5126447677612305, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "bdf43e03d12027e26301fde6547acaa0e9b2f9d02b250c958506947c0a71663d", "metric_row_sha256": "76d9713274034bd613d5513c2fb85ac96a1a49be9591ce61f3989e60ae1afde2", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.1438177448749673, "expected_voiced_frames": 319, "missing_voiced_frames": 9 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.02821316614420063, "expected_voiced_frames": 319, "missing_voiced_frames": 9 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.9249565601348877 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "6786194c0c460344d01a8d12083542ddfbf10dff2c22b9c292cc9df018abf47d", "metric_row_sha256": "6eb6289382030ce00ad6beb16a174f653182d911929267dbda8324525ab3a45c", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.1321596786652695, "expected_voiced_frames": 322, "missing_voiced_frames": 11 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.034161490683229816, "expected_voiced_frames": 322, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.437601327896118 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.025515293400340866, "unvoiced_miss_rate": -0.27077651717344264, "automatic_mos": -0.17394304275512695, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": true, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": true, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "3c1d754ad393a05d23c60c508d6ac929d022537166915a94fcfa385693f4fa65", "metric_row_sha256": "1b1c5ef1070d211e312d1ebf892bb519c8b89815d49b5564cd4a99605ce8fb0f", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.0540918709295957, "expected_voiced_frames": 319, "missing_voiced_frames": 10 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03134796238244514, "expected_voiced_frames": 319, "missing_voiced_frames": 10 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.1362924575805664 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "b96f0846f2b56b8f9785769ba8039796aa91ab7f084477d3f8fb99bcef74217d", "metric_row_sha256": "0011936c20ba257cbd83027c7575bbf55846a192c153ff73ef72dd4bcfd81e71", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.0796071643299365, "expected_voiced_frames": 323, "missing_voiced_frames": 11 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.034055727554179564, "expected_voiced_frames": 323, "missing_voiced_frames": 11 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3102355003356934 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.0034419747656664335, "unvoiced_miss_rate": 0.01953144073672565, "automatic_mos": -0.43587589263916016, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "3c1d754ad393a05d23c60c508d6ac929d022537166915a94fcfa385693f4fa65", "metric_row_sha256": "1b1c5ef1070d211e312d1ebf892bb519c8b89815d49b5564cd4a99605ce8fb0f", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.0540918709295957, "expected_voiced_frames": 319, "missing_voiced_frames": 10 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03134796238244514, "expected_voiced_frames": 319, "missing_voiced_frames": 10 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.1362924575805664 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "beb16365d019943c464287cd31fe8c4428621dd677f3eee639f4aab8616ab695", "metric_row_sha256": "7a408cb4b8214cb8aae44418856efdc7c54cae25233e11bba91702a1be537948", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.057533845695262, "expected_voiced_frames": 321, "missing_voiced_frames": 10 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03115264797507788, "expected_voiced_frames": 321, "missing_voiced_frames": 10 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5721683502197266 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.060031089732325515, "unvoiced_miss_rate": 0.029206176132712605, "automatic_mos": -0.49326062202453613, "recognition": null }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null }, "wer": { "candidate": null, "reference": null, "candidate_minus_reference_output_rate": null } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "3c1d754ad393a05d23c60c508d6ac929d022537166915a94fcfa385693f4fa65", "metric_row_sha256": "1b1c5ef1070d211e312d1ebf892bb519c8b89815d49b5564cd4a99605ce8fb0f", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.0540918709295957, "expected_voiced_frames": 319, "missing_voiced_frames": 10 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03134796238244514, "expected_voiced_frames": 319, "missing_voiced_frames": 10 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.1362924575805664 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "bff6e00207b265cf7d4bc6cab4d1d21c8cc9ac08e896a483e5c82368ace56637", "metric_row_sha256": "ecee139e9b8429fb038570e85ef2e3853c2e93b1cb997a6b965d9eab18b8d0ab", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.114122960661921, "expected_voiced_frames": 322, "missing_voiced_frames": 10 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.031055900621118012, "expected_voiced_frames": 322, "missing_voiced_frames": 10 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6295530796051025 }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "detail": null } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": 0.22558331489562988, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "64142a0704ebcd7e46a990da2b7126f3d88884e793f5287548fa9fe90b6c4d45", "metric_row_sha256": "d6f667de504cd84ca6b5482ca1ed7fea592bace84e6b17634fc26b73ae7a496f", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7355642318725586 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "80bf274d7883bff30b22f1eae1ea2a40cb7d3f7fdc0bc75162dc90e44faa6cc8", "metric_row_sha256": "93af44f7a57626ff76d175f707d8f8afafff620b97986756b0c2277c36aa0126", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5099809169769287 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": 0.03388571739196777, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "64142a0704ebcd7e46a990da2b7126f3d88884e793f5287548fa9fe90b6c4d45", "metric_row_sha256": "d6f667de504cd84ca6b5482ca1ed7fea592bace84e6b17634fc26b73ae7a496f", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7355642318725586 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "dcd1ac6a72f272d6fc7c1fa80f103ef2d5b3ce3e2829883bc3a331f60018de8b", "metric_row_sha256": "7117ec5e8ed38b26f831b7ab41495452d4a46fa3f4f237c9b8565ea8303da8fc", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.701678514480591 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": 0.04519510269165039, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "64142a0704ebcd7e46a990da2b7126f3d88884e793f5287548fa9fe90b6c4d45", "metric_row_sha256": "d6f667de504cd84ca6b5482ca1ed7fea592bace84e6b17634fc26b73ae7a496f", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7355642318725586 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "949f626d7910d668b396b77782f8be4f4609bfac31d94289026258b8bb674406", "metric_row_sha256": "12b5e2175ed336e36e4024cd2a7afab57b52af5d16df5856a413fee685fd76c2", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.690369129180908 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.019737958908081055, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "850e24c72c52e6d61b67acd28c07ccf83d6d2d38708626df49176dc638b4079e", "metric_row_sha256": "f448cab89a31517ce48f980245518cbce59abb4d2fe907b33bb67ab7dfb25997", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.657235622406006 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "d72575fda7a9e8db365de1c65a8e9b355296ef98e347a1c3226524f9593f4159", "metric_row_sha256": "f1a7e424123148be7529629f74c834bdb9e486d3e4fd02cb40583fd2766884c3", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.676973581314087 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": 0.009043216705322266, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "850e24c72c52e6d61b67acd28c07ccf83d6d2d38708626df49176dc638b4079e", "metric_row_sha256": "f448cab89a31517ce48f980245518cbce59abb4d2fe907b33bb67ab7dfb25997", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.657235622406006 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "4be53edf46fd57f9609b6ce92a1aab70cdacb467ad6343ff715c09f767c2a743", "metric_row_sha256": "d74fefeba5b811e55c33fb8efc2ad85dfaab2f4bb64ddf1fe7a8d601387362da", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6481924057006836 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": null, "unvoiced_miss_rate": null, "automatic_mos": -0.018623828887939453, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": true, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "850e24c72c52e6d61b67acd28c07ccf83d6d2d38708626df49176dc638b4079e", "metric_row_sha256": "f448cab89a31517ce48f980245518cbce59abb4d2fe907b33bb67ab7dfb25997", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.657235622406006 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "36fdb8b277d1d4bbbecea19a534132c8cf66a9df5d90aad10e00c30fbc1497f3", "metric_row_sha256": "f3e93f1610f7c74e4a17d1ffb9552b5b557e7255c7cb3d8c375a1c5610159bc9", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6758594512939453 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.060384186317260724, "unvoiced_miss_rate": -0.7280139778683747, "automatic_mos": 0.06164193153381348, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "3c8cce327674d6b982384dbe1c1023abee975ad80c53e236e5c7c9673377e060", "metric_row_sha256": "1d9a12cd9215bc70bafb785e108d309d58b7a4a530a90c87683f9121cd6dc4a5", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.1992891404406865, "expected_voiced_frames": 101, "missing_voiced_frames": 26 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25742574257425743, "expected_voiced_frames": 101, "missing_voiced_frames": 26 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6110925674438477 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "5f8ac38b5291f8debac169ad194b32c2528998f3b21c63de76ea4fcad1970559", "metric_row_sha256": "98759df4e3be417aec0bfa924e0d4d5460549575a56cce1882de0dc641a8cfd3", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.259673326757947, "expected_voiced_frames": 102, "missing_voiced_frames": 27 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2647058823529412, "expected_voiced_frames": 102, "missing_voiced_frames": 27 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.549450635910034 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.013564278845002775, "unvoiced_miss_rate": 0, "automatic_mos": -0.12583589553833008, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "3c8cce327674d6b982384dbe1c1023abee975ad80c53e236e5c7c9673377e060", "metric_row_sha256": "1d9a12cd9215bc70bafb785e108d309d58b7a4a530a90c87683f9121cd6dc4a5", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.1992891404406865, "expected_voiced_frames": 101, "missing_voiced_frames": 26 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25742574257425743, "expected_voiced_frames": 101, "missing_voiced_frames": 26 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6110925674438477 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "4996340607c4cc19fc9ac2b9c3af1076d309a22c436c13e51e71de4be9774c34", "metric_row_sha256": "340d1503d6fbe6d190e18cc9aabfd986f31b0eaa12b172da5fd6dda4aeb0ad1a", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.2128534192856892, "expected_voiced_frames": 101, "missing_voiced_frames": 26 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25742574257425743, "expected_voiced_frames": 101, "missing_voiced_frames": 26 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7369284629821777 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.0033027059022194827, "unvoiced_miss_rate": 0, "automatic_mos": -0.1238565444946289, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "3c8cce327674d6b982384dbe1c1023abee975ad80c53e236e5c7c9673377e060", "metric_row_sha256": "1d9a12cd9215bc70bafb785e108d309d58b7a4a530a90c87683f9121cd6dc4a5", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.1992891404406865, "expected_voiced_frames": 101, "missing_voiced_frames": 26 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25742574257425743, "expected_voiced_frames": 101, "missing_voiced_frames": 26 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6110925674438477 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "4813534b07e440e59d21780658c3e525df5c694ba786c2940867140d6113e5d0", "metric_row_sha256": "78e56f0dda4d4ffa7565039757205dde69d1d811a37180726948d5e77003d9c7", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.195986434538467, "expected_voiced_frames": 101, "missing_voiced_frames": 26 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25742574257425743, "expected_voiced_frames": 101, "missing_voiced_frames": 26 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7349491119384766 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.24870127824828803, "unvoiced_miss_rate": -0.25645722659827785, "automatic_mos": -0.05065608024597168, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "457f31848de2fa59f93ec0d0a0e9af6af47a98d36342215e91dda378224f9246", "metric_row_sha256": "ad60e166242da020758f9037eccdc348ee655290900b0b9dd94f452bbdbdc616", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.219665073820892, "expected_voiced_frames": 103, "missing_voiced_frames": 25 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.24271844660194175, "expected_voiced_frames": 103, "missing_voiced_frames": 25 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6067724227905273 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "b1c6a99b922e64084c3bcbd428cc3fce45a43dd63410c1ccc7b2189d87daa79c", "metric_row_sha256": "8e340e8a55cebb7813b7d188eceb38304ac2e0053578ead141c241bc41c28a55", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.46836635206918, "expected_voiced_frames": 106, "missing_voiced_frames": 26 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.24528301886792453, "expected_voiced_frames": 106, "missing_voiced_frames": 26 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.657428503036499 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.020810739437036485, "unvoiced_miss_rate": -0.7281553398058249, "automatic_mos": -0.14687800407409668, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "457f31848de2fa59f93ec0d0a0e9af6af47a98d36342215e91dda378224f9246", "metric_row_sha256": "ad60e166242da020758f9037eccdc348ee655290900b0b9dd94f452bbdbdc616", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.219665073820892, "expected_voiced_frames": 103, "missing_voiced_frames": 25 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.24271844660194175, "expected_voiced_frames": 103, "missing_voiced_frames": 25 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6067724227905273 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "ec944cdc276ea33c8b75991ecfa6164d17c52ca1ccf47b9ddae76f197078e162", "metric_row_sha256": "ae4dc783da1a483b5a026f45da028d26a798bb44b20f81075cf62e5151018a4a", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.2404758132579286, "expected_voiced_frames": 104, "missing_voiced_frames": 26 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25, "expected_voiced_frames": 104, "missing_voiced_frames": 26 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.753650426864624 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": -0.017182596444283327, "unvoiced_miss_rate": -0.7281553398058249, "automatic_mos": -0.11943268775939941, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "457f31848de2fa59f93ec0d0a0e9af6af47a98d36342215e91dda378224f9246", "metric_row_sha256": "ad60e166242da020758f9037eccdc348ee655290900b0b9dd94f452bbdbdc616", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.219665073820892, "expected_voiced_frames": 103, "missing_voiced_frames": 25 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.24271844660194175, "expected_voiced_frames": 103, "missing_voiced_frames": 25 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6067724227905273 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "3ab764980d96f338357e648a94fd46a5c0d8fb8e3b014786b4dbd5b4503240a5", "metric_row_sha256": "ec9a8cbb534f9fa59a7b588d534960a19f8c0091329121e92687980e0cc55c10", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.2368476702651754, "expected_voiced_frames": 104, "missing_voiced_frames": 26 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25, "expected_voiced_frames": 104, "missing_voiced_frames": 26 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7262051105499268 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.0019915342503709077, "unvoiced_miss_rate": -0.9706853038245, "automatic_mos": 0.1579606533050537, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "4e42dc407b50b4a18aaab6a4427849647c46b4a87f40a8d6bdb328d29722b70c", "metric_row_sha256": "d6540a4521657738d08c2bc39db10dbaf84c7321bbdc558bf38a8a74b400ed50", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.495310864822625, "expected_voiced_frames": 101, "missing_voiced_frames": 1 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.009900990099009901, "expected_voiced_frames": 101, "missing_voiced_frames": 1 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2838759422302246 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "853d3bf3d3d2c31934f366de3a4411d89890cbb2bdf551b5e388dd024fc612ef", "metric_row_sha256": "6a72cd685dbeb7606aa16a5048a6647dabfc24ae49a24f5712bc680de9f74718", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.493319330572254, "expected_voiced_frames": 102, "missing_voiced_frames": 2 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.0196078431372549, "expected_voiced_frames": 102, "missing_voiced_frames": 2 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.125915288925171 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.014904335576169991, "unvoiced_miss_rate": 0, "automatic_mos": 0.07360029220581055, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "4e42dc407b50b4a18aaab6a4427849647c46b4a87f40a8d6bdb328d29722b70c", "metric_row_sha256": "d6540a4521657738d08c2bc39db10dbaf84c7321bbdc558bf38a8a74b400ed50", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.495310864822625, "expected_voiced_frames": 101, "missing_voiced_frames": 1 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.009900990099009901, "expected_voiced_frames": 101, "missing_voiced_frames": 1 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2838759422302246 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "67a0bf877283e452676529b80fd0e8169f52c5ddc9d3e95a8fc3e3a309fa7276", "metric_row_sha256": "fc2f8a20366d822927eb47d63083d621c84902e5efbab8816e9213acbf732b87", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.480406529246455, "expected_voiced_frames": 101, "missing_voiced_frames": 1 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.009900990099009901, "expected_voiced_frames": 101, "missing_voiced_frames": 1 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.210275650024414 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 0, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.01735161465108659, "unvoiced_miss_rate": 0, "automatic_mos": 0.025321483612060547, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "4e42dc407b50b4a18aaab6a4427849647c46b4a87f40a8d6bdb328d29722b70c", "metric_row_sha256": "d6540a4521657738d08c2bc39db10dbaf84c7321bbdc558bf38a8a74b400ed50", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.495310864822625, "expected_voiced_frames": 101, "missing_voiced_frames": 1 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.009900990099009901, "expected_voiced_frames": 101, "missing_voiced_frames": 1 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2838759422302246 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "a21ddb6e0f662806c84fd2d89b10a14e1cc382f76091a49a9470c2a3a7a88e3f", "metric_row_sha256": "2f1e6c9ca49e5cc7867332e683295ee52681e3647e6d0915cd6d20b554794630", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.4779592501715384, "expected_voiced_frames": 101, "missing_voiced_frames": 1 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.009900990099009901, "expected_voiced_frames": 101, "missing_voiced_frames": 1 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.258554458618164 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "PML", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.1537904011259097, "unvoiced_miss_rate": -3.6361971056970135, "automatic_mos": -0.1801009178161621, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": { "automatic_mos": true, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_flags_applied_only_to_PML": true, "harm_observed": true, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "3c88ce3a9611c3d258608d755cf632e48d1fa9bc24735aeee23fc6d108cedea7", "metric_row_sha256": "ee888a493809a802b405eb7d3ecbae9965a36f9c918578465469884d1067266a", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.594640072805242, "expected_voiced_frames": 103, "missing_voiced_frames": 5 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04854368932038835, "expected_voiced_frames": 103, "missing_voiced_frames": 5 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.0839598178863525 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "PML", "waveform_sha256": "53ef5dc779c5e814bb541bdd59ea0c0fb48f663ebd0e3b0af516119e62eff5e9", "metric_row_sha256": "18bcb76a4c0e7294f4f3e16966a15299c7b011d1e183c0ac3994fa49969a434f", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.440849671679332, "expected_voiced_frames": 106, "missing_voiced_frames": 9 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.08490566037735849, "expected_voiced_frames": 106, "missing_voiced_frames": 9 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2640607357025146 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_SHAPE", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.08249046704259522, "unvoiced_miss_rate": 0.04667662434652692, "automatic_mos": -0.013791799545288086, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "3c88ce3a9611c3d258608d755cf632e48d1fa9bc24735aeee23fc6d108cedea7", "metric_row_sha256": "ee888a493809a802b405eb7d3ecbae9965a36f9c918578465469884d1067266a", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.594640072805242, "expected_voiced_frames": 103, "missing_voiced_frames": 5 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04854368932038835, "expected_voiced_frames": 103, "missing_voiced_frames": 5 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.0839598178863525 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_SHAPE", "waveform_sha256": "66464f8e5e7f66373ebae514cb028af4a75cf8897a3ebf87b1857623e6e1a250", "metric_row_sha256": "c7b3b050a9ed6c9dd50b2bd0528ec540900ca49e8c73baba427603e7f2270ff2", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.5121496057626467, "expected_voiced_frames": 104, "missing_voiced_frames": 5 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04807692307692308, "expected_voiced_frames": 104, "missing_voiced_frames": 5 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.0977516174316406 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_POSTCOLOR_M1", "reference_variant": "RAW_FULL", "pair_status": "recorded", "candidate_minus_reference": { "relative_f0_mae_cents": 0.02873754686079577, "unvoiced_miss_rate": 0.04667662434652692, "automatic_mos": -0.12401103973388672, "recognition": 0 }, "unvoiced_miss_delta_unit": "percentage_point", "recognition": { "cer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 }, "wer": { "candidate": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "reference": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "candidate_minus_reference_output_rate": 0 } }, "recognition_harm_threshold": null, "harm_flags": {}, "harm_flags_applied_only_to_PML": false, "harm_observed": false, "required_observations_missing": false, "candidate": { "row_status": "recorded", "variant": "RAW_POSTCOLOR_M1", "waveform_sha256": "3c88ce3a9611c3d258608d755cf632e48d1fa9bc24735aeee23fc6d108cedea7", "metric_row_sha256": "ee888a493809a802b405eb7d3ecbae9965a36f9c918578465469884d1067266a", "evaluation_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.594640072805242, "expected_voiced_frames": 103, "missing_voiced_frames": 5 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04854368932038835, "expected_voiced_frames": 103, "missing_voiced_frames": 5 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.0839598178863525 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } }, "reference": { "row_status": "recorded", "variant": "RAW_FULL", "waveform_sha256": "4355e40ee25065c7b2f2fc11f03578ec8b22387c92e76ed345121d2ad778c3e0", "metric_row_sha256": "19c7cb54e2ff9be3abdf53e7afdd727d5e270227a8282007c2c71a34787deb9a", "evaluation_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.565902525944446, "expected_voiced_frames": 104, "missing_voiced_frames": 5 }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04807692307692308, "expected_voiced_frames": 104, "missing_voiced_frames": 5 }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2079708576202393 }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0, "detail": { "status": "ok", "value": 0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0, "output_rate": 0, "delta": 0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0, "output_rate": 0, "delta": 0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" } } } } } ], "descriptive_stats": [ { "reference_variant": "PML", "metric": "automatic_mos", "available_pairs": 30, "planned_pairs": 30, "mean": -0.014812278747558593, "min": -0.4072532653808594, "max": 0.29399681091308594, "improved_pairs": 15, "worsened_pairs": 15, "equal_pairs": 0, "fixed_PML_harm_pairs": 10 }, { "reference_variant": "PML", "metric": "relative_f0_mae_cents", "available_pairs": 20, "planned_pairs": 30, "mean": 0.16439950552814353, "min": -0.24870127824828803, "max": 3.599069484496858, "improved_pairs": 13, "worsened_pairs": 7, "equal_pairs": 0, "fixed_PML_harm_pairs": 0 }, { "reference_variant": "PML", "metric": "unvoiced_miss_rate", "available_pairs": 20, "planned_pairs": 30, "mean": -0.42540948573385995, "min": -3.6361971056970135, "max": 1.7094017094017089, "improved_pairs": 14, "worsened_pairs": 5, "equal_pairs": 1, "fixed_PML_harm_pairs": 0 }, { "reference_variant": "PML", "metric": "recognition", "available_pairs": 12, "planned_pairs": 30, "mean": 0, "min": 0, "max": 0, "improved_pairs": 0, "worsened_pairs": 0, "equal_pairs": 12, "fixed_PML_harm_pairs": null }, { "reference_variant": "RAW_SHAPE", "metric": "automatic_mos", "available_pairs": 30, "planned_pairs": 30, "mean": -0.07944854100545247, "min": -0.5261802673339844, "max": 0.2291431427001953, "improved_pairs": 12, "worsened_pairs": 18, "equal_pairs": 0, "fixed_PML_harm_pairs": null }, { "reference_variant": "RAW_SHAPE", "metric": "relative_f0_mae_cents", "available_pairs": 20, "planned_pairs": 30, "mean": -0.33947423673548033, "min": -4.873747182363205, "max": 0.17884948626291663, "improved_pairs": 13, "worsened_pairs": 7, "equal_pairs": 0, "fixed_PML_harm_pairs": null }, { "reference_variant": "RAW_SHAPE", "metric": "unvoiced_miss_rate", "available_pairs": 20, "planned_pairs": 30, "mean": -0.03409240335485231, "min": -1.7094017094017089, "max": 2.5641025641025634, "improved_pairs": 7, "worsened_pairs": 8, "equal_pairs": 5, "fixed_PML_harm_pairs": null }, { "reference_variant": "RAW_SHAPE", "metric": "recognition", "available_pairs": 12, "planned_pairs": 30, "mean": 0.004166666666666666, "min": 0, "max": 0.04999999999999999, "improved_pairs": 0, "worsened_pairs": 1, "equal_pairs": 11, "fixed_PML_harm_pairs": null }, { "reference_variant": "RAW_FULL", "metric": "automatic_mos", "available_pairs": 30, "planned_pairs": 30, "mean": -0.11417097250620524, "min": -0.5126447677612305, "max": 0.2528233528137207, "improved_pairs": 7, "worsened_pairs": 23, "equal_pairs": 0, "fixed_PML_harm_pairs": null }, { "reference_variant": "RAW_FULL", "metric": "relative_f0_mae_cents", "available_pairs": 20, "planned_pairs": 30, "mean": -0.30582079745153157, "min": -4.459810555870776, "max": 0.02873754686079577, "improved_pairs": 14, "worsened_pairs": 6, "equal_pairs": 0, "fixed_PML_harm_pairs": null }, { "reference_variant": "RAW_FULL", "metric": "unvoiced_miss_rate", "available_pairs": 20, "planned_pairs": 30, "mean": -0.055927447076982106, "min": -1.7094017094017089, "max": 1.7094017094017089, "improved_pairs": 4, "worsened_pairs": 8, "equal_pairs": 8, "fixed_PML_harm_pairs": null }, { "reference_variant": "RAW_FULL", "metric": "recognition", "available_pairs": 12, "planned_pairs": 30, "mean": 0, "min": 0, "max": 0, "improved_pairs": 0, "worsened_pairs": 0, "equal_pairs": 12, "fixed_PML_harm_pairs": null } ], "per_pitch_recording_equal_weight_deltas": [ { "reference_variant": "PML", "pitch_semitones": 0, "metric": "relative_f0_mae_cents", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null } ], "contributing_recordings": 0, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": null, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "PML", "pitch_semitones": 0, "metric": "unvoiced_miss_rate", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null } ], "contributing_recordings": 0, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": null, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "PML", "pitch_semitones": 0, "metric": "automatic_mos", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.16736245155334473 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.09642565250396729 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.16779863834381104 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.3193577527999878 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.10292267799377441 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.024089121818542482, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "PML", "pitch_semitones": 0, "metric": "recognition", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 } ], "contributing_recordings": 2, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": 0, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "PML", "pitch_semitones": -3, "metric": "relative_f0_mae_cents", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.04230436612243982 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.009602170046984826 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 1.7709958293097685 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.12676540861383812 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.15454273228277438 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": 0.2875562304487463, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "PML", "pitch_semitones": -3, "metric": "unvoiced_miss_rate", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.537078189946709 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.3758146615289476 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.5908584169453737 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.1523724487320091 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.4922356022333263 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.21484058789858954, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "PML", "pitch_semitones": -3, "metric": "automatic_mos", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.022080659866333008 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.23590624332427979 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.1180189847946167 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.25223350524902344 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.0054929256439208984 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": 0.025853061676025392, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "PML", "pitch_semitones": -3, "metric": "recognition", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 } ], "contributing_recordings": 2, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": 0, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "PML", "pitch_semitones": 3, "metric": "relative_f0_mae_cents", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.08431824617049588 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.08681244391649545 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.12103980034134931 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.004788937262214743 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.0778909676881403 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": 0.04124278060754079, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "PML", "pitch_semitones": 3, "metric": "unvoiced_miss_rate", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.052042434902785706 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -1.9886901345867907 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 1.6425120772946853 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.5823150906955753 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -2.3034412047607566 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.6359783835691303, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "PML", "pitch_semitones": 3, "metric": "automatic_mos", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.02687227725982666 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.25934457778930664 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.2155524492263794 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.29059815406799316 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.0110701322555542 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.04620077610015869, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "PML", "pitch_semitones": 3, "metric": "recognition", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 } ], "contributing_recordings": 2, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": 0, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_SHAPE", "pitch_semitones": 0, "metric": "relative_f0_mae_cents", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null } ], "contributing_recordings": 0, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": null, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_SHAPE", "pitch_semitones": 0, "metric": "unvoiced_miss_rate", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null } ], "contributing_recordings": 0, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": null, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_SHAPE", "pitch_semitones": 0, "metric": "automatic_mos", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.05512869358062744 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.010830879211425781 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.12288045883178711 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.3492159843444824 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.02146446704864502 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.07693448066711425, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_SHAPE", "pitch_semitones": 0, "metric": "recognition", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 } ], "contributing_recordings": 2, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": 0, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_SHAPE", "pitch_semitones": -3, "metric": "relative_f0_mae_cents", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.032405065546574185 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.006258500527055455 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.6673517569330163 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.15625237497660538 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.01718750914101963 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.173387641214032, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_SHAPE", "pitch_semitones": -3, "metric": "unvoiced_miss_rate", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.05912836122839438 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -1.2820512820512817 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.5657415254018818 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.36407766990291246 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.2042518130647836, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_SHAPE", "pitch_semitones": -3, "metric": "automatic_mos", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.024490714073181152 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.0624467134475708 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.02279794216156006 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.24453222751617432 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.13635694980621338 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.07314622402191162, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_SHAPE", "pitch_semitones": -3, "metric": "recognition", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.024999999999999994 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 } ], "contributing_recordings": 2, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": 0.012499999999999997, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_SHAPE", "pitch_semitones": 3, "metric": "relative_f0_mae_cents", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.05399737870741261 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.09543034513028537 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -2.619056098435294 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.001121569418395696 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.048697401309382604 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.5055608322569286, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_SHAPE", "pitch_semitones": 3, "metric": "unvoiced_miss_rate", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.22073328702137163 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.6993006993006996 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 1.7094017094017089 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.13237100347750635 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.02333831217326346 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": 0.13606700635507896, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_SHAPE", "pitch_semitones": 3, "metric": "automatic_mos", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.03905308246612549 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.2244551181793213 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.17560279369354248 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.48102807998657227 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.02990424633026123 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.08826491832733155, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_SHAPE", "pitch_semitones": 3, "metric": "recognition", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 } ], "contributing_recordings": 2, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": 0, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_FULL", "pitch_semitones": 0, "metric": "relative_f0_mae_cents", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null } ], "contributing_recordings": 0, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": null, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_FULL", "pitch_semitones": 0, "metric": "unvoiced_miss_rate", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null } ], "contributing_recordings": 0, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": null, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_FULL", "pitch_semitones": 0, "metric": "automatic_mos", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.009454488754272461 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.03283393383026123 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.15326297283172607 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.3581966161727905 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.013285636901855469 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.10431067943572998, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_FULL", "pitch_semitones": 0, "metric": "recognition", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 } ], "contributing_recordings": 2, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": 0, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_FULL", "pitch_semitones": -3, "metric": "relative_f0_mae_cents", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.030070797919501846 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 4.4195860149365274e-07 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.49415939342141635 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.13378629204432713 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.006939945271031922 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.13299119733953516, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_FULL", "pitch_semitones": -3, "metric": "unvoiced_miss_rate", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.17283237797872247 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.8547008547008544 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.11828501333748728 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.36407766990291246 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.25466517784900045, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_FULL", "pitch_semitones": -3, "metric": "automatic_mos", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.028446316719055176 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.018187522888183594 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.054284095764160156 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.26018786430358887 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.12164461612701416 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.08927507400512695, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_FULL", "pitch_semitones": -3, "metric": "recognition", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 } ], "contributing_recordings": 2, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": 0, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_FULL", "pitch_semitones": 3, "metric": "relative_f0_mae_cents", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.058066230936161034 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 1.378544877272958e-05 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -2.334057611324879 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.02418651176131381 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.02304458075594118 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.478650397563528, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_FULL", "pitch_semitones": 3, "metric": "unvoiced_miss_rate", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.11882539048616615 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.8547008547008544 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.28281313888510307 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.02333831217326346 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": 0.1428102836950362, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_FULL", "pitch_semitones": 3, "metric": "automatic_mos", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.16666781902313232 }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0.18914353847503662 }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.21481406688690186 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.5029526948928833 }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": -0.049344778060913086 } ], "contributing_recordings": 5, "planned_recordings": 5, "all_five_recordings_available": true, "available_recording_equal_weight_mean": -0.1489271640777588, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" }, { "reference_variant": "RAW_FULL", "pitch_semitones": 3, "metric": "recognition", "per_recording": [ { "case": "librispeech_1089-134686-0008", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1188-133604-0003", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_121-121726-0012", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 }, { "case": "librispeech_1221-135766-0005", "available_seed_pairs": 0, "expected_seed_pairs": 2, "mean": null }, { "case": "librispeech_1089-134686-0032", "available_seed_pairs": 2, "expected_seed_pairs": 2, "mean": 0 } ], "contributing_recordings": 2, "planned_recordings": 5, "all_five_recordings_available": false, "available_recording_equal_weight_mean": 0, "definition": "Average the 2 paired seeds within recording, then equal weight available recordings", "statistical_independence": "5 recordings / 4 speakers; seeds are paired within-recording repeats" } ], "scientific_source_hashes": { "source_summary_sha256": "ba6b98e9ab6ecc4be1e9c71d5ae7c6ead062e33980fe928510a361a3ab42ff07", "completion_receipt_sha256": "18c0d248a2d988f4002a1b486f9295ef1f6ee8fcdfa606315d4c4f30f8589819", "original_90_summary_sha256": "1bee17075fc78e939cb4e36fe10ff23e5be87ff12585bf9758f8e6c0136b7f14", "candidate_source_freeze_sha256": "ae5e2ab11825397ec61fcf2e2764e93fa0b30488ac46a28eec8b4b9cdf0ea390", "old_metric_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf" }, "automatic_mos_by_variant_unranked": [ { "variant": "PML", "available_conditions": 30, "recordings": 5, "per_recording_6condition_means": [ { "case": "librispeech_1089-134686-0008", "conditions": 6, "mean": 3.31997021039327 }, { "case": "librispeech_1188-133604-0003", "conditions": 6, "mean": 3.3413516680399575 }, { "case": "librispeech_121-121726-0012", "conditions": 6, "mean": 3.0144155025482178 }, { "case": "librispeech_1221-135766-0005", "conditions": 6, "mean": 3.5264798402786255 }, { "case": "librispeech_1089-134686-0032", "conditions": 6, "mean": 3.463968276977539 } ], "recording_equal_weight_mean": 3.333237099647522, "per_pitch_recording_equal_weight_means": [ { "pitch_semitones": 0, "recordings": 5, "paired_seeds_per_recording": 2, "mean": 3.36033833026886 }, { "pitch_semitones": -3, "recordings": 5, "paired_seeds_per_recording": 2, "mean": 3.446730446815491 }, { "pitch_semitones": 3, "recordings": 5, "paired_seeds_per_recording": 2, "mean": 3.1926425218582155 } ], "definition_ja": "各録音の3音高×2seedを等重み平均し、5録音を等重み平均。全条件が揃うため全30値平均と一致。採用順位・総合指標ではない。" }, { "variant": "RAW_POSTCOLOR_M1", "available_conditions": 30, "recordings": 5, "per_recording_6condition_means": [ { "case": "librispeech_1089-134686-0008", "conditions": 6, "mean": 3.3920753399531045 }, { "case": "librispeech_1188-133604-0003", "conditions": 6, "mean": 3.538577159245809 }, { "case": "librispeech_121-121726-0012", "conditions": 6, "mean": 2.9259714682896933 }, { "case": "librispeech_1221-135766-0005", "conditions": 6, "mean": 3.2390833695729575 }, { "case": "librispeech_1089-134686-0032", "conditions": 6, "mean": 3.496416767438253 } ], "recording_equal_weight_mean": 3.3184248208999634, "per_pitch_recording_equal_weight_means": [ { "pitch_semitones": 0, "recordings": 5, "paired_seeds_per_recording": 2, "mean": 3.3362492084503175 }, { "pitch_semitones": -3, "recordings": 5, "paired_seeds_per_recording": 2, "mean": 3.472583508491516 }, { "pitch_semitones": 3, "recordings": 5, "paired_seeds_per_recording": 2, "mean": 3.1464417457580565 } ], "definition_ja": "各録音の3音高×2seedを等重み平均し、5録音を等重み平均。全条件が揃うため全30値平均と一致。採用順位・総合指標ではない。" }, { "variant": "RAW_SHAPE", "available_conditions": 30, "recordings": 5, "per_recording_6condition_means": [ { "case": "librispeech_1089-134686-0008", "conditions": 6, "mean": 3.3948803742726645 }, { "case": "librispeech_1188-133604-0003", "conditions": 6, "mean": 3.439332922299703 }, { "case": "librispeech_121-121726-0012", "conditions": 6, "mean": 3.0330651998519897 }, { "case": "librispeech_1221-135766-0005", "conditions": 6, "mean": 3.5973421335220337 }, { "case": "librispeech_1089-134686-0032", "conditions": 6, "mean": 3.5247461795806885 } ], "recording_equal_weight_mean": 3.397873361905416, "per_pitch_recording_equal_weight_means": [ { "pitch_semitones": 0, "recordings": 5, "paired_seeds_per_recording": 2, "mean": 3.4131836891174316 }, { "pitch_semitones": -3, "recordings": 5, "paired_seeds_per_recording": 2, "mean": 3.545729732513428 }, { "pitch_semitones": 3, "recordings": 5, "paired_seeds_per_recording": 2, "mean": 3.234706664085388 } ], "definition_ja": "各録音の3音高×2seedを等重み平均し、5録音を等重み平均。全条件が揃うため全30値平均と一致。採用順位・総合指標ではない。" }, { "variant": "RAW_FULL", "available_conditions": 30, "recordings": 5, "per_recording_6condition_means": [ { "case": "librispeech_1089-134686-0008", "conditions": 6, "mean": 3.453961888949076 }, { "case": "librispeech_1188-133604-0003", "conditions": 6, "mean": 3.4804114500681558 }, { "case": "librispeech_121-121726-0012", "conditions": 6, "mean": 3.0667585134506226 }, { "case": "librispeech_1221-135766-0005", "conditions": 6, "mean": 3.6128624280293784 }, { "case": "librispeech_1089-134686-0032", "conditions": 6, "mean": 3.54898468653361 } ], "recording_equal_weight_mean": 3.4325957934061684, "per_pitch_recording_equal_weight_means": [ { "pitch_semitones": 0, "recordings": 5, "paired_seeds_per_recording": 2, "mean": 3.4405598878860473 }, { "pitch_semitones": -3, "recordings": 5, "paired_seeds_per_recording": 2, "mean": 3.561858582496643 }, { "pitch_semitones": 3, "recordings": 5, "paired_seeds_per_recording": 2, "mean": 3.2953689098358154 } ], "definition_ja": "各録音の3音高×2seedを等重み平均し、5録音を等重み平均。全条件が揃うため全30値平均と一致。採用順位・総合指標ではない。" } ] }