{
  "rubric_id": "humanness-annotator-audio-review-v1",
  "n_recordings": 50,
  "languages": {
    "en": 45,
    "de": 5
  },
  "model_versions": [
    "gemini-3.8-flash"
  ],
  "results": {
    "composite": {
      "n": 50,
      "mean": 3.2630000000000003,
      "bootstrap_95_ci": [
        3.1919750000000002,
        3.3320000000000003
      ],
      "recordings_rated_0_to_2": 0
    },
    "global_caption_fidelity": {
      "n": 50,
      "mean": 3.82,
      "bootstrap_95_ci": [
        3.7,
        3.92
      ],
      "recordings_rated_0_to_2": 0
    },
    "global_emotion_fit": {
      "n": 50,
      "mean": 3.74,
      "bootstrap_95_ci": [
        3.6,
        3.86
      ],
      "recordings_rated_0_to_2": 1
    },
    "sentence_emotion_fit": {
      "n": 50,
      "mean": 2.78,
      "bootstrap_95_ci": [
        2.62,
        2.92
      ],
      "recordings_rated_0_to_2": 14
    },
    "utterance_talking_style_fit": {
      "n": 50,
      "mean": 3.86,
      "bootstrap_95_ci": [
        3.76,
        3.94
      ],
      "recordings_rated_0_to_2": 0
    },
    "sentence_delivery_fit": {
      "n": 50,
      "mean": 3.1,
      "bootstrap_95_ci": [
        3.02,
        3.2
      ],
      "recordings_rated_0_to_2": 0
    },
    "burst_description_fit": {
      "n": 50,
      "mean": 3.02,
      "bootstrap_95_ci": [
        2.86,
        3.16
      ],
      "recordings_rated_0_to_2": 4
    },
    "burst_timing_fit": {
      "n": 50,
      "mean": 2.52,
      "bootstrap_95_ci": [
        2.36,
        2.68
      ],
      "recordings_rated_0_to_2": 24
    },
    "event_precision": {
      "n": 50,
      "mean": 2.88,
      "bootstrap_95_ci": [
        2.74,
        3.0
      ],
      "recordings_rated_0_to_2": 6
    },
    "event_coverage": {
      "n": 50,
      "mean": 3.46,
      "bootstrap_95_ci": [
        3.32,
        3.6
      ],
      "recordings_rated_0_to_2": 0
    },
    "transcript_fidelity": {
      "n": 50,
      "mean": 4.3,
      "bootstrap_95_ci": [
        4.14,
        4.46
      ],
      "recordings_rated_0_to_2": 0
    },
    "speech_timing_fit": {
      "n": 50,
      "mean": 3.42,
      "bootstrap_95_ci": [
        3.28,
        3.56
      ],
      "recordings_rated_0_to_2": 1
    },
    "readability": {
      "n": 50,
      "mean": 3.86,
      "bootstrap_95_ci": [
        3.76,
        3.94
      ],
      "recordings_rated_0_to_2": 0
    }
  },
  "local_segment_reviews": 662,
  "local_segment_diagnostics": {
    "emotion_fit_mean": 3.06797583081571,
    "delivery_fit_mean": 3.3836858006042294,
    "aggregation": "Unweighted across local segments; diagnostic only, excluded from the clip-weighted composite"
  },
  "estimated_list_price_usd": 0.43679025,
  "usage_totals": {
    "promptTokenCount": 334502,
    "candidatesTokenCount": 49577,
    "thoughtsTokenCount": 0,
    "totalTokenCount": 384079
  },
  "pricing": {
    "date": "2026-10-11",
    "input_usd_per_million": 0.75,
    "output_usd_per_million": 3.75,
    "source": "https://ai.google.dev/gemini-api/docs/pricing",
    "not_an_invoice": true
  },
  "rubric": {
    "id": "humanness-annotator-audio-review-v1",
    "scale": "Integer 0\u20135: 0 unusable/wrong; 1 major errors; 2 partly supported; 3 useful with notable errors; 4 accurate with minor errors; 5 fully correct and useful",
    "composite_weights": {
      "global_caption_fidelity": 0.15,
      "global_emotion_fit": 0.1,
      "sentence_emotion_fit": 0.2,
      "utterance_talking_style_fit": 0.15,
      "sentence_delivery_fit": 0.1,
      "burst_description_fit": 0.1,
      "burst_timing_fit": 0.1,
      "event_precision": 0.05,
      "event_coverage": 0.05
    },
    "definitions": {
      "global_caption_fidelity": "Does GENERAL describe the audible voice, performance and environment accurately, without invented demographic/identity/psychological claims?",
      "global_emotion_fit": "Are the whole-utterance emotion descriptions supported by audible affect, including intensity and changes? Important omissions reduce usefulness; restrained descriptions are appropriate when affect is ambiguous.",
      "sentence_emotion_fit": "Do emotion cues fit their particular spoken segments and intensity? Consider both false emotion tags and important missing affect. No emotion tag is not a claim of neutrality: do not penalize appropriate restraint on ambiguous segments.",
      "utterance_talking_style_fit": "Do GENERAL's speaking-style claims match audible tempo, articulation, breathiness, pitch movement, register, emphasis and performance style?",
      "sentence_delivery_fit": "Do local delivery/voice tags fit the corresponding spoken segment rather than merely repeating a global trait?",
      "burst_description_fit": "Does each free-form burst caption match the actual nonverbal vocal sound in its selected interval? Reward acoustic accuracy, not verbosity.",
      "burst_timing_fit": "Do absolute event starts/ends cover the audible events without unnecessary unrelated speech?",
      "event_precision": "Are proposed intervals real nonverbal vocal events rather than spoken syllables or spurious detections? This is a reviewer rating, not measured detection precision.",
      "event_coverage": "Are salient audible nonverbal events covered? This is a reviewer rating, not measured detection recall.",
      "transcript_fidelity": "Does the recognized word sequence match what is heard? This is diagnostic, not part of the weighted composite.",
      "speech_timing_fit": "Do speech budgets and pause intervals fit the audio? This is diagnostic, separate from mechanical formatting validation.",
      "readability": "Is the annotation understandable, concise and useful? This is diagnostic, not part of the weighted composite."
    },
    "aggregation": "Equal weight per recording; fixed weights above. Per-segment ratings are diagnostic and do not overweight recordings with more sentences.",
    "empty_event_policy": "If no event is audible and none is proposed, event ratings are 5. False proposals or omitted audible events reduce the corresponding ratings.",
    "important_limitations": "Single automated reviewer; 50 reused synthetic acting recordings, 45 English/5 German; no human reference labels, no inter-rater agreement. New rubric is not comparable numerically to the older MOSS/procedural composite. No settings are optimized using these new ratings."
  },
  "automated_reviewer_not_human_ground_truth": true,
  "bootstrap": "10,000 clip resamples, seed 20261011; excludes reviewer bias and domain shift",
  "retuning_using_these_scores": false,
  "new_rubric_not_comparable_to_previous_composite": true
}