{
  "method": "temperature_scaling",
  "temperature": 0.565484,
  "fit_split": "validation",
  "fit_manifest": "data/manifests/splits/val_manifest.jsonl",
  "test_manifest": "data/manifests/splits/test_manifest.jsonl",
  "apply_at_inference": "probs = softmax(logits / temperature)",
  "note": "Temperature scaling changes probabilities/confidence, not argmax class predictions.",
  "validation_raw": {
    "accuracy": 0.970439,
    "f1_macro": 0.957292,
    "f1_weighted": 0.970565,
    "nll": 0.238713,
    "brier": 0.074267,
    "ece_15_bins": 0.129845,
    "mean_confidence": 0.840781
  },
  "validation_calibrated": {
    "accuracy": 0.970439,
    "f1_macro": 0.957292,
    "f1_weighted": 0.970565,
    "nll": 0.136917,
    "brier": 0.049335,
    "ece_15_bins": 0.009176,
    "mean_confidence": 0.969347
  },
  "test_raw": {
    "accuracy": 0.96663,
    "f1_macro": 0.954043,
    "f1_weighted": 0.966779,
    "nll": 0.248412,
    "brier": 0.078886,
    "ece_15_bins": 0.126878,
    "mean_confidence": 0.839925
  },
  "test_calibrated": {
    "accuracy": 0.96663,
    "f1_macro": 0.954043,
    "f1_weighted": 0.966779,
    "nll": 0.152494,
    "brier": 0.055011,
    "ece_15_bins": 0.009874,
    "mean_confidence": 0.968902
  }
}