{
  "model": "TRMSelector \u2014 recursive PASS/FAIL chunk gate (skip head)",
  "params": 1905921,
  "arch": {
    "hidden": 256,
    "H_cycles": 3,
    "L_cycles": 4,
    "head_dropout": 0.1
  },
  "encoder": "openai/text-embedding-ada-002",
  "emb_dim": 1539,
  "threshold": 0.942,
  "lr": 0.000113,
  "metrics": {
    "threshold": 0.942,
    "micro_precision": 0.9345,
    "micro_recall": 0.6624,
    "micro_f1": 0.7753,
    "macro_f1": 0.502,
    "exact_set_match": 0.2,
    "avg_selected": 3.73,
    "avg_true_pass": 5.27,
    "num_groups": 45,
    "best_epoch": 53
  },
  "bench": {
    "gold_easy": {
      "groups": 9,
      "precision": 0.8571,
      "recall": 0.2069,
      "f1": 0.3333
    },
    "gold_medium": {
      "groups": 2,
      "precision": 1.0,
      "recall": 0.9211,
      "f1": 0.9589
    },
    "gold_hard": {
      "groups": 5,
      "precision": 1.0,
      "recall": 0.9667,
      "f1": 0.9831
    },
    "llm_held_out": {
      "groups": 29,
      "precision": 0.7436,
      "recall": 0.3625,
      "f1": 0.4874
    }
  },
  "data": {
    "train_groups": 1184,
    "gold_train_groups": 48,
    "test_groups": 45,
    "gold_test_groups": 16,
    "train_candidates": 27544,
    "train_pass": 5567,
    "train_fail": 21977,
    "pass_ratio": 0.202,
    "avg_candidates_per_train_group": 23.3,
    "avg_pass_per_train_group": 4.7
  },
  "trained_at": "2026-07-07T02:33:14",
  "source": "https://github.com/s3777091/recursive_models"
}