{ "schema_version": 2, "sample_id": "aishell4_test_m_r003s01c01", "evaluation_track": "moss_x1_dev_rocm_native_whole_context", "process_status": "succeeded", "engineering_completeness_status": "engineering_complete", "accuracy_ranking_allowed": true, "gate": { "transcript_coverage_status": "passed", "generation_termination_status": "coverage_complete_stop_reason_uninstrumented", "raw_tail_status": "passed_complete_parse", "repeat_timeline_status": "passed", "speaker_continuity_status": "structure_present_accuracy_scored" }, "coverage": { "audio_duration_seconds": 2282.21, "reference_last_voice_boundary_seconds": 2247.8433, "tail_tolerance_seconds": 1.0, "prediction_first_start_seconds": 4.15, "prediction_last_end_seconds": 2247.84, "coverage_ratio_to_reference_last_voice": 0.9999985319261356, "trailing_gap_to_reference_last_voice_seconds": 0.0032999999998537533, "trailing_gap_to_audio_end_seconds": 34.36999999999989, "passed": true }, "raw_parser_audit": { "raw_nonempty": true, "raw_character_count": 17410, "raw_ending": "，往下边儿都铺好铺好那个地毯那类的[2224.84][2224.94][S05]呃最好让孩子就是说假如说孩子真是摔下来了，也不归着磕着，下边儿铺好地毯，做好防护吧，因为冬天冷了[2232.64][2233.14][S01]所以呃这个安全这块儿，你们下边儿都注意一下，呃那今天就到这儿，你们就是该把自己工作做的做好，呃把这个孩子的安全也要放到第一位[2245.04][2245.24][S05]嗯[2246.14][2246.04][S01]啊好了，今天会就到此[2247.84]\n", "result_segment_count": 381, "parser_segment_count": 381, "start_speaker_header_count": 381, "parser_header_mismatch_indices": [], "bracket_balance_open_minus_close": 0, "raw_roundtrip_matches_parser_serialization": true, "raw_ends_on_complete_parsed_segment": true }, "generation": { "generated_tokens": 13462, "requested_max_new_tokens": 65536, "equals_requested_max_new_tokens": false, "equals_known_prior_observed_16384_ceiling": false, "suspected_token_ceiling": false, "eos_observed": null, "stop_reason": null, "instrumentation_limitation": "runner_did_not_retain_token_ids_eos_or_stopping_criterion" }, "timeline": { "numeric_segment_count": 381, "negative_start_indices": [], "non_positive_duration_indices": [], "start_regression_indices": [], "end_regression_indices_allowed_nested_overlap": [ 50, 69, 82, 87, 95, 102, 107, 155, 160, 163, 167, 170, 172, 178, 183, 185, 191, 200, 204, 213, 219, 222, 224, 230, 232, 236, 281, 285, 291, 293, 295, 300, 302, 304, 309, 312, 315, 318, 323, 325, 330, 336, 350, 362, 364, 366, 368, 371, 374 ], "out_of_audio_bounds_indices": [], "adjacent_exact_speaker_text_duplicate_indices": [ 40, 156, 175, 225, 369 ], "severe_repetition_detected": false, "longform_diagnostics": { "schema_version": 1, "protocol": { "name": "longform-engineering-diagnostics", "version": "v1", "accuracy_reference_required": false, "repetition_check": "exact_adjacent_normalized_token_blocks", "repetition_block_sizes_tokens": [ 12, 24, 48, 96 ] }, "nonempty_output": true, "text_character_count": 9295, "normalized_token_count": 8345, "characters_per_audio_minute": 244.368397, "runtime": { "audio_duration_seconds": 2282.21, "inference_seconds": 3780.029454813004, "realtime_factor": 1.6563022047984208, "job_id": null }, "repetition": { "adjacent_duplicate_block_counts": { "12": 0, "24": 0, "48": 0, "96": 0 }, "maximum_detected_duplicate_block_tokens": 0 }, "timeline": { "available": true, "segment_count": 381, "timed_segment_count": 381, "first_start_seconds": 4.15, "last_end_seconds": 2247.84, "trailing_unsegmented_seconds": 34.37, "invalid_span_count": 0, "start_regression_count": 0, "adjacent_identical_text_count": 5 } } }, "speaker_structure": { "speaker_count": 6, "labels": [ "S01", "S02", "S03", "S04", "S05", "S06" ], "segment_count_by_label": { "S01": 128, "S02": 27, "S03": 11, "S04": 90, "S05": 45, "S06": 80 } }, "metrics": { "status": "formal_complete_sample", "accuracy_ranking_allowed": true, "transcription": { "primary_metric": "cer", "cer": { "rate": 0.256856, "substitutions": 670, "deletions": 1680, "insertions": 179, "reference_tokens": 9846 }, "punctuation": { "precision": 0.770175, "recall": 0.839388, "f1": 0.803294, "matches": 439, "reference_count": 523, "hypothesis_count": 570 }, "terms": { "reference_count": 0, "hits": 0, "recall": 1.0 } }, "speaker_text": { "valid": true, "invalid_reason": "", "cer": { "rate": 0.155687, "errors": 1340, "reference_characters": 8607, "hypothesis_characters": 8345 }, "cpcer": { "rate": 0.135471, "errors": 1166, "reference_characters": 8607, "hypothesis_characters": 8345 }, "delta_cp": -0.020216, "reference_speakers": 6, "hypothesis_speakers": 6, "speaker_assignment": [ { "reference_index": 0, "hypothesis_index": 0, "errors": 546 }, { "reference_index": 1, "hypothesis_index": 1, "errors": 14 }, { "reference_index": 2, "hypothesis_index": 2, "errors": 15 }, { "reference_index": 3, "hypothesis_index": 3, "errors": 234 }, { "reference_index": 4, "hypothesis_index": 4, "errors": 136 }, { "reference_index": 5, "hypothesis_index": 5, "errors": 221 } ] }, "diarization": { "protocol": { "collar_seconds": 0.25 }, "overlap_included": { "der": 0.049479, "confusion": 11.5362, "false_alarm": 30.9499, "missed_detection": 53.6372, "total": 1942.7069 }, "overlap_excluded": { "der": 0.035303, "confusion": 6.6352, "false_alarm": 29.9712, "missed_detection": 26.3511, "total": 1783.3616 }, "jer": 0.055774, "reference_speaker_count": 6, "hypothesis_speaker_count": 6, "speaker_count_error": 0 }, "timestamps": { "status": "diagnostic_native_segment_pairing_not_token_timestamp_parity", "accuracy_ranking_allowed": false, "metric": { "schema_version": 1, "protocol": { "text_alignment": "rapidfuzz_levenshtein_equal_blocks", "aggregation": "unique_matched_reference_hypothesis_segment_pairs", "percentile_interpolation": "linear", "tolerances_seconds": [ 0.25, 0.5, 1.0 ] }, "valid": true, "reference_segment_count": 641, "hypothesis_segment_count": 381, "reference_token_count": 9846, "hypothesis_token_count": 8345, "matched_token_count": 7496, "reference_token_match_coverage": 0.761324, "matched_segment_pair_count": 572, "start_absolute_error_seconds": { "count": 572, "mean": 2.987186, "median": 1.17, "p90": 8.615743, "p95": 10.089535, "max": 16.77798 }, "end_absolute_error_seconds": { "count": 572, "mean": 3.08055, "median": 1.597893, "p90": 8.403108, "p95": 10.525835, "max": 18.89454 }, "time_iou": { "count": 572, "mean": 0.467835, "median": 0.393506, "p90": 0.947608, "p95": 0.976786, "max": 0.996876 }, "tolerance_hits": { "0.25s": { "start_hit_rate": 0.437063, "end_hit_rate": 0.40035, "both_hit_rate": 0.197552 }, "0.5s": { "start_hit_rate": 0.45979, "end_hit_rate": 0.440559, "both_hit_rate": 0.236014 }, "1s": { "start_hit_rate": 0.486014, "end_hit_rate": 0.456294, "both_hit_rate": 0.251748 } }, "reference_timeline_diagnostics": { "negative_start_count": 0, "non_positive_duration_count": 0, "start_regression_count": 0 }, "hypothesis_timeline_diagnostics": { "negative_start_count": 0, "non_positive_duration_count": 0, "start_regression_count": 0 } } } }, "runtime": { "model_load_seconds": 2.496422451993567, "inference_seconds": 3780.029454813004, "rtf_excluding_model_load": 1.6563022047984208, "runtime_metrics_valid": false, "runtime_comparability": "invalid_external_cpu_contention", "contamination": { "schema_version": 1, "sample_id": "aishell4_test_m_r003s01c01", "contamination_type": "external_cpu_test_suite_contention", "observed_at_approx": "2026-07-21T18:36:00+08:00/2026-07-21T18:37:00+08:00", "duration_seconds_approx": 20, "workload": "project-wide ruff and pytest invoked outside the inference orchestrator", "impact": { "accuracy_artifacts_valid": true, "runtime_metrics_valid": false, "runtime_comparability_valid": false, "rerun_for_speed_authorized": false }, "handling": "Keep the original inference for accuracy; exclude M timing and RTF from runtime rankings." } }, "evidence": [ { "path": "runs/moss-0.9b-meeting-first-pass-v2-prepared-20260721/02-aishell4_test_m_r003s01c01/inference/result.json", "size_bytes": 1249, "sha256": "92bb526b0f6d1a132aa899ff6027f9352c54efb915985dfe2dccaf2f767034da" }, { "path": "runs/moss-0.9b-meeting-first-pass-v2-prepared-20260721/02-aishell4_test_m_r003s01c01/inference/raw/transcript.txt", "size_bytes": 35240, "sha256": "845ed6267e102c099e9c12ad8b269b3d4f64799be1acd713eadf2d10e3188a29" }, { "path": "runs/moss-0.9b-meeting-first-pass-v2-prepared-20260721/02-aishell4_test_m_r003s01c01/inference/segments.json", "size_bytes": 64640, "sha256": "db9fd15defc21de7d670ca45f135d50b1af902d83946576e70b5ebf90c60ac2a" }, { "path": "runs/moss-0.9b-meeting-first-pass-v2-prepared-20260721/02-aishell4_test_m_r003s01c01/execution-snapshot/request.source.json", "size_bytes": 3699, "sha256": "14bfc8fa75b428023211d78f7be98005faf531d34e9988860860be8f9e298759" }, { "path": "data/samples/meetings/aishell-4/M_R003S01C01/audio.wav", "size_bytes": 73030798, "sha256": "724df9fc8f4210359f6c2296932f4a07db78121bde85bded7044074209d3b483" }, { "path": "data/samples/meetings/aishell-4/M_R003S01C01/reference.txt", "size_bytes": 29226, "sha256": "b3c42f5b2a1853ba959ce9c34f5dfc088e2686794f90b9ddd03fc6a0b266c5ab" }, { "path": "data/samples/meetings/aishell-4/M_R003S01C01/reference.speaker.txt", "size_bytes": 47870, "sha256": "343eee920cd4d2f3bd2e0490e717172030c5817fb012626871d2e8973ea5ff69" }, { "path": "data/samples/meetings/aishell-4/M_R003S01C01/reference.segments.jsonl", "size_bytes": 185777, "sha256": "3687e60aace55005894d02de3cb1358c9823010c00dd681dbbd7cd6ce84f55bf" }, { "path": "data/samples/meetings/aishell-4/M_R003S01C01/reference.rttm", "size_bytes": 42175, "sha256": "bf67a4df2f2d2ae78cccfeefc47ca4244db043e6bbc7a44f83dc9b3d3ee3af57" }, { "path": "runs/moss-0.9b-meeting-first-pass-v2-prepared-20260721/02-aishell4_test_m_r003s01c01/runtime-contamination.json", "size_bytes": 616, "sha256": "51cd5e39f8c0e0023dbd136e306f7585109d6160efeaa7a7aa9c60c813594ab5" } ] }