eval_results.json

{
  "checkpoint": "value_function_model",
  "n_samples": 3000,
  "accuracy": 0.82,
  "precision": 0.8553868402024585,
  "recall": 0.9441340782122905,
  "f1": 0.8975720789074355,
  "confusion_matrix": {
    "tp": 2366,
    "fp": 400,
    "tn": 94,
    "fn": 140
  },
  "calibration": {
    "avg_prob_for_positive": 0.9026213149047335,
    "avg_prob_for_negative": 0.7740231474884126
  }
}