{
  "schema_version": 4,
  "split": "test",
  "benchmark": "natural",
  "population_trials": 6000,
  "frozen_execution": {
    "status": "COMPLETE",
    "execution_count": 1,
    "post_test_tuning": false,
    "validation_status": "READY_TO_SHARE"
  },
  "metric_notes": {
    "wer": "Uncapped utterance-level config-consistent Whisper large-v3 WER; insertions can make WER exceed 100%.",
    "a_sw": "Study-defined acoustic-switch indicator.",
    "utmos": "Non-intrusive predicted naturalness; it does not establish target-speaker or lexical correctness.",
    "dnsmos_p808": "DNSMOS P.808 is displayed at corpus level; per-case P.808 scores are not shown in these listening examples.",
    "estoi": "Intrusive intelligibility estimate against the clean target."
  },
  "selection_protocol": {
    "common_filter": "Trial identities and waveforms are unchanged from the frozen listening-case selection; displayed transcripts and WER are re-audited consistently with Whisper large-v3.",
    "evidence_repair": "Wrong-speaker BSRNN-TSE case retained to expose candidate repair; large-v3 scores BSRNN-TSE/BSRNN-TSE-CDCS-5/Qwen-TSE-fixed-CSG at 144.44/11.11/0% WER.",
    "speaker_grounding": "Ungrounded acoustic-switch collapse retained; fixed and adaptive CSG restore target-side speaker identity and lower large-v3 WER from 100% to 25%.",
    "content_grounding": "Former lexical-drift player set retained as a listening control because large-v3 assigns every displayed output the exact target transcript.",
    "quality_grounding": "Matched 0%-WER outputs retained to expose the fidelity–predicted-naturalness operating-point difference."
  },
  "cases": [
    {
      "slug": "evidence-repair",
      "trial_id": "test:5142-33396-0010_8555-292519-0012:8555",
      "target_text": "through the black night rain he sang to me",
      "systems": {
        "bsrnn_tse": {
          "output_text": "in the stern i carved the tail up almost as high as the head",
          "wer": 1.444444,
          "a_sw": true,
          "speaker_margin": -0.527178,
          "utmos": 3.386915,
          "estoi": -0.014471
        },
        "bsrnn_tse_cdcs_5": {
          "output_text": "through the black night rain he sang to",
          "wer": 0.111111,
          "a_sw": false,
          "speaker_margin": 0.39393,
          "utmos": 2.620063,
          "estoi": 0.78723
        },
        "qwen_tse_ud": {
          "output_text": "through the black night rain he sang to me",
          "wer": 0,
          "a_sw": false,
          "speaker_margin": 0.262971,
          "utmos": 3.425434,
          "estoi": 0.66927
        },
        "qwen_tse_fixed_csg": {
          "output_text": "through the black night rain he sang to me",
          "wer": 0,
          "a_sw": false,
          "speaker_margin": 0.249747,
          "utmos": 3.650644,
          "estoi": 0.692586
        },
        "qwen_tse_adaptive_csg": {
          "output_text": "through the black night rain he sang to me",
          "wer": 0.0,
          "a_sw": false,
          "speaker_margin": 0.2431435417884214,
          "utmos": 3.33864164352417,
          "estoi": 0.6969271193161576,
          "wer_counts": {
            "substitutions": 0,
            "deletions": 0,
            "insertions": 0,
            "reference_words": 9
          }
        }
      },
      "audio_assets": [
        {
          "role": "qwen_tse_adaptive_csg",
          "path": "public/assets/audio/noisy/evidence-repair/qwen-adaptive-csg.wav",
          "sha256": "af514a32372c22cfaf4cfdc43b3b806dab44d54c5100ef8f007f8734745f09c0",
          "sample_rate_hz": 24000,
          "duration_seconds": 3.48
        }
      ]
    },
    {
      "slug": "speaker-grounding",
      "trial_id": "test:1995-1826-0015_908-157963-0005:1995",
      "target_text": "she had almost forgotten that it was here within touch and sight",
      "systems": {
        "bsrnn_tse": {
          "output_text": "she had almost forgotten that it was here within touch and sight",
          "wer": 0,
          "a_sw": false,
          "speaker_margin": 0.54507,
          "utmos": 3.300329,
          "estoi": 0.869006
        },
        "bsrnn_tse_cdcs_5": {
          "output_text": "she had almost forgotten that it was here he can touch and sight",
          "wer": 0.166667,
          "a_sw": false,
          "speaker_margin": 0.561049,
          "utmos": 3.324447,
          "estoi": 0.864569
        },
        "qwen_tse_ud": {
          "output_text": "you",
          "wer": 1,
          "a_sw": true,
          "speaker_margin": -0.008244,
          "utmos": 2.976541,
          "estoi": 0.003788
        },
        "qwen_tse_fixed_csg": {
          "output_text": "she had almost forgotten that it was here they didnt touch and sigh",
          "wer": 0.25,
          "a_sw": false,
          "speaker_margin": 0.673998,
          "utmos": 3.858996,
          "estoi": 0.747898
        },
        "qwen_tse_adaptive_csg": {
          "output_text": "she had almost forgotten that it was here they didnt touch and sigh",
          "wer": 0.25,
          "a_sw": false,
          "speaker_margin": 0.6698132990792056,
          "utmos": 3.923950672149658,
          "estoi": 0.7556042568834386,
          "wer_counts": {
            "substitutions": 2,
            "deletions": 0,
            "insertions": 1,
            "reference_words": 12
          }
        }
      },
      "audio_assets": [
        {
          "role": "qwen_tse_adaptive_csg",
          "path": "public/assets/audio/noisy/speaker-grounding/qwen-adaptive-csg.wav",
          "sha256": "0a421c9120d19589886fed6f3c8ce5b6006919f36959394ed7b7596bb91c4171",
          "sample_rate_hz": 24000,
          "duration_seconds": 3.56
        }
      ]
    },
    {
      "slug": "content-grounding",
      "trial_id": "test:260-123440-0013_8230-279154-0016:8230",
      "target_text": "in actual fact there are doubtless various",
      "systems": {
        "bsrnn_tse": {
          "output_text": "in actual fact there are doubtless various",
          "wer": 0,
          "a_sw": false,
          "speaker_margin": 0.477253,
          "utmos": 2.918162,
          "estoi": 0.647725
        },
        "bsrnn_tse_cdcs_5": {
          "output_text": "in actual fact there are doubtless various",
          "wer": 0,
          "a_sw": false,
          "speaker_margin": 0.491468,
          "utmos": 2.922394,
          "estoi": 0.651861
        },
        "qwen_tse_ud": {
          "output_text": "in actual fact there are doubtless various",
          "wer": 0,
          "a_sw": false,
          "speaker_margin": 0.459699,
          "utmos": 3.596734,
          "estoi": 0.31403
        },
        "qwen_tse_fixed_csg": {
          "output_text": "in actual fact there are doubtless various",
          "wer": 0,
          "a_sw": false,
          "speaker_margin": 0.421932,
          "utmos": 3.556462,
          "estoi": 0.608632
        },
        "qwen_tse_adaptive_csg": {
          "output_text": "in actual fact there are doubtless various",
          "wer": 0.0,
          "a_sw": false,
          "speaker_margin": 0.4201470881115929,
          "utmos": 3.6364340782165527,
          "estoi": 0.6003693106932193,
          "wer_counts": {
            "substitutions": 0,
            "deletions": 0,
            "insertions": 0,
            "reference_words": 7
          }
        }
      },
      "audio_assets": [
        {
          "role": "qwen_tse_adaptive_csg",
          "path": "public/assets/audio/noisy/content-grounding/qwen-adaptive-csg.wav",
          "sha256": "99a5ab25f42f7e7adc1dba5ed3b5492cb64c67ec014dff0353218c670eacac1b",
          "sample_rate_hz": 24000,
          "duration_seconds": 3.48
        }
      ]
    },
    {
      "slug": "quality-grounding",
      "trial_id": "test:3729-6852-0019_5639-40744-0037:3729",
      "target_text": "i tell him to give me some coffee if it is good",
      "systems": {
        "bsrnn_tse": {
          "output_text": "i tell him to give me some coffee if it is good",
          "wer": 0,
          "a_sw": false,
          "speaker_margin": 0.530555,
          "utmos": 1.479945,
          "estoi": 0.736379
        },
        "bsrnn_tse_cdcs_5": {
          "output_text": "i tell him to give me some coffee if it is good",
          "wer": 0,
          "a_sw": false,
          "speaker_margin": 0.486536,
          "utmos": 1.306923,
          "estoi": 0.692086
        },
        "qwen_tse_ud": {
          "output_text": "i tell him to give me some coffee if it is good",
          "wer": 0,
          "a_sw": false,
          "speaker_margin": 0.562963,
          "utmos": 3.780135,
          "estoi": 0.626619
        },
        "qwen_tse_fixed_csg": {
          "output_text": "i tell him to give me some coffee if it is good",
          "wer": 0,
          "a_sw": false,
          "speaker_margin": 0.560803,
          "utmos": 3.827171,
          "estoi": 0.620886
        },
        "qwen_tse_adaptive_csg": {
          "output_text": "i tell him to give me some coffee if it is good",
          "wer": 0.0,
          "a_sw": false,
          "speaker_margin": 0.558653760509469,
          "utmos": 3.84786057472229,
          "estoi": 0.618567535094135,
          "wer_counts": {
            "substitutions": 0,
            "deletions": 0,
            "insertions": 0,
            "reference_words": 12
          }
        }
      },
      "audio_assets": [
        {
          "role": "qwen_tse_adaptive_csg",
          "path": "public/assets/audio/noisy/quality-grounding/qwen-adaptive-csg.wav",
          "sha256": "c36a7554768fab857efede9ab10d04144562b6ebaa66ae1bacbcf30c949571b8",
          "sample_rate_hz": 24000,
          "duration_seconds": 3.32
        }
      ]
    }
  ],
  "asr": {
    "model": "openai/whisper-large-v3",
    "revision": "06f233fe06e710322aca913c1bc4249a0d71fce1",
    "task": "English/transcribe",
    "decoding": "greedy",
    "repetition_penalty": 1.1,
    "batch_size": 32
  },
  "reference_protocol": "Config-consistent WER against the clean-target transcript produced by the same Whisper large-v3 configuration.",
  "adaptive_addition": {
    "date": "2026-10-25",
    "source": "Existing frozen Natural Noisy TEST adaptive-CSG waveforms and completed post-hoc large-v3/RP=1.10 audit.",
    "matching": "Exact trial IDs and target/mixture SHA-256 match the existing four examples.",
    "new_inference": false,
    "added_audio_asset_hashes": "Each case.audio_assets lists the newly added adaptive-CSG waveform."
  }
}