continuation_benchmark/validation_eval/summary.json

{
  "version": "validation_eval_score_v1",
  "generated_utc": "2026-05-03T10:40:48.528789Z",
  "n_plasmids": 80,
  "n_comp_windows": 1397,
  "n_surp_windows": 413,
  "per_stratum_completion": [
    {
      "stratum": "A_only",
      "n": 30,
      "Base_mean": -6.1205,
      "SFT_mean": -5.2073,
      "RL_mean": -7.6367,
      "RL-SFT": -2.4293,
      "RL-Base": -1.5162,
      "median_RL-SFT": -2.341,
      "RL_gt_SFT": 0,
      "RL_gt_Base": 2,
      "pct_RL_gt_SFT": 0.0,
      "pct_RL_gt_Base": 6.7
    },
    {
      "stratum": "B_only",
      "n": 25,
      "Base_mean": -7.0608,
      "SFT_mean": -5.5863,
      "RL_mean": -8.1659,
      "RL-SFT": -2.5796,
      "RL-Base": -1.1051,
      "median_RL-SFT": -2.6391,
      "RL_gt_SFT": 0,
      "RL_gt_Base": 4,
      "pct_RL_gt_SFT": 0.0,
      "pct_RL_gt_Base": 16.0
    },
    {
      "stratum": "D_intersection",
      "n": 25,
      "Base_mean": -6.9151,
      "SFT_mean": -6.1175,
      "RL_mean": -7.8682,
      "RL-SFT": -1.7508,
      "RL-Base": -0.9531,
      "median_RL-SFT": -1.6148,
      "RL_gt_SFT": 2,
      "RL_gt_Base": 3,
      "pct_RL_gt_SFT": 8.0,
      "pct_RL_gt_Base": 12.0
    }
  ],
  "per_stratum_surprisal": [
    {
      "stratum": "A_only",
      "n": 30,
      "Base_mean": -13.3006,
      "SFT_mean": -7.4583,
      "RL_mean": -10.9045,
      "RL-SFT": -3.4462,
      "RL-Base": 2.3961,
      "median_RL-SFT": -3.5137,
      "RL_gt_SFT": 0,
      "RL_gt_Base": 30,
      "pct_RL_gt_SFT": 0.0,
      "pct_RL_gt_Base": 100.0
    },
    {
      "stratum": "B_only",
      "n": 25,
      "Base_mean": -14.1787,
      "SFT_mean": -8.4321,
      "RL_mean": -11.6772,
      "RL-SFT": -3.2451,
      "RL-Base": 2.5015,
      "median_RL-SFT": -3.4234,
      "RL_gt_SFT": 0,
      "RL_gt_Base": 25,
      "pct_RL_gt_SFT": 0.0,
      "pct_RL_gt_Base": 100.0
    },
    {
      "stratum": "D_intersection",
      "n": 25,
      "Base_mean": -13.3266,
      "SFT_mean": -7.9818,
      "RL_mean": -10.8727,
      "RL-SFT": -2.8908,
      "RL-Base": 2.454,
      "median_RL-SFT": -2.8733,
      "RL_gt_SFT": 0,
      "RL_gt_Base": 25,
      "pct_RL_gt_SFT": 0.0,
      "pct_RL_gt_Base": 100.0
    }
  ],
  "overall": {
    "comp_RL_minus_SFT_mean": -2.264244972006103,
    "comp_RL_minus_SFT_median": -2.3358309253969005,
    "comp_n_RL_gt_SFT": 2,
    "surp_RL_minus_SFT_mean": -3.2098014044445575,
    "surp_RL_minus_SFT_median": -3.2402958245504467,
    "surp_n_RL_gt_SFT": 0
  }
}