{
  "ci95": {
    "groups": 91,
    "iterations": 2000,
    "level": 0.95,
    "method": "whole-group percentile bootstrap; fixed full label denominator",
    "metrics": {
      "accuracy": {
        "high": 0.5915750915750916,
        "low": 0.5183150183150184,
        "median": 0.554945054945055
      },
      "macro_f1": {
        "high": 0.5792936660243523,
        "low": 0.5024747832145906,
        "median": 0.5408729979177656
      },
      "uar": {
        "high": 0.5915750915750916,
        "low": 0.5183150183150184,
        "median": 0.554945054945055
      }
    },
    "seed": 20260817,
    "small_group_warning": false
  },
  "claim_scope": "Recomputes supplied predictions; does not rerun or authenticate the claimed model",
  "environment": {
    "dependencies": "Python standard library only",
    "implementation": "CPython",
    "python": "3.14.6",
    "scorer_sha256": "1a38d5ddba9a2b3106f85926e48b9195d8c9e6e82b56ecb6630a21ceeedbbe22"
  },
  "inference_executed": false,
  "inference_source_sha256": "646b70de8c5ffeee378f65184ffe3017dd384822ad45357d0ff6330217fa5ff6",
  "input_manifest_sha256": "960149e3b798955f4db4d2bea42f373e30d923371abb4e2dfb89e055687fd4d7",
  "input_verification": {
    "audio_bytes": 39917840,
    "audio_seconds": 1246.68175,
    "audio_verified": true,
    "examples": 546,
    "preprocessing": "none"
  },
  "labels": [
    "anger",
    "disgust",
    "fear",
    "happiness",
    "neutral",
    "sadness"
  ],
  "metrics": {
    "abstentions": 2,
    "accuracy": 0.5567765567765568,
    "coverage": 0.9963369963369964,
    "macro_f1": 0.5424321519607127,
    "n": 546,
    "per_class": {
      "anger": {
        "f1": 0.702928870292887,
        "precision": 0.5675675675675675,
        "recall": 0.9230769230769231,
        "support": 91
      },
      "disgust": {
        "f1": 0.5566037735849056,
        "precision": 0.48760330578512395,
        "recall": 0.6483516483516484,
        "support": 91
      },
      "fear": {
        "f1": 0.4766355140186916,
        "precision": 0.4146341463414634,
        "recall": 0.5604395604395604,
        "support": 91
      },
      "happiness": {
        "f1": 0.3333333333333333,
        "precision": 0.5365853658536586,
        "recall": 0.24175824175824176,
        "support": 91
      },
      "neutral": {
        "f1": 0.6962025316455697,
        "precision": 0.8208955223880597,
        "recall": 0.6043956043956044,
        "support": 91
      },
      "sadness": {
        "f1": 0.4888888888888889,
        "precision": 0.75,
        "recall": 0.3626373626373626,
        "support": 91
      }
    },
    "uar": 0.5567765567765568,
    "weighted_f1": 0.5424321519607127
  },
  "mode": "offline_supplied_prediction_replay",
  "model_identity_independently_verified": false,
  "model_provenance": {
    "id": "oruk-fourier",
    "inference_description": "One bounded current /v1/audio/emotions call per attempted input; no retries; argmax over all returned selected emotion scores, fixed spelling aliases, lexical tie break; out-of-taxonomy winner/empty output abstains; no probability renormalization",
    "revision": null,
    "revision_evidence": null
  },
  "prediction_document_claims_inference_executed": true,
  "predictions_sha256": "cd0c4b477a49106db3ddbe628e8594afe2dfac663b8646dcdacc6c38e8b771c7",
  "purpose": "evaluation",
  "reference_semantics": "Intended acted emotion encoded in upstream filename, not listener-judgment labels or an independent annotation.",
  "schema": "oruk.emotion-replay/1",
  "training_overlap": "unknown; same-source diagnostic only, no held-out or generalization claim"
}