{
  "arms": [
    {
      "base_zero_subset": {
        "behavior_diversity_mean": 0.7777777777777778,
        "candidate_count_mean": 3.625,
        "functional_diversity_mean": 0.5208333333333334,
        "hidden_pass_candidates_mean": 0.0,
        "parse_success_mean": 3.375
      },
      "behavior_diversity_rate_mean": 0.75625,
      "candidate_count_mean": 3.45,
      "coverage": 0.7,
      "coverage_pct": 70.0,
      "false_repair_count": 0,
      "false_repair_rate": 0.0,
      "functional_diversity_rate_mean": 0.4770833333333333,
      "hidden_pass_candidates_mean": 1.875,
      "incremental_forward_tokens": 69645,
      "key": "base_k4",
      "name": "Base K4",
      "parse_success_mean": 3.3625,
      "recovered": [],
      "remaining": [
        15,
        16,
        22,
        26,
        31,
        34,
        35,
        36,
        39,
        42,
        43,
        44,
        48,
        55,
        59,
        60,
        67,
        70,
        73,
        77,
        81,
        83,
        84,
        87
      ],
      "total_forward_tokens": 69645,
      "visible_repair_pass_count": 0,
      "zero_to_one": 0,
      "zero_to_one_rate": 0.0
    },
    {
      "base_zero_subset": {
        "behavior_diversity_mean": 0.4899577835323017,
        "candidate_count_mean": 26.375,
        "functional_diversity_mean": 0.12257363738009504,
        "hidden_pass_candidates_mean": 1.125,
        "parse_success_mean": 24.5
      },
      "behavior_diversity_rate_mean": 0.6699040017263572,
      "candidate_count_mean": 10.275,
      "coverage": 0.8,
      "coverage_pct": 80.0,
      "false_repair_count": 0,
      "false_repair_rate": 0.0,
      "functional_diversity_rate_mean": 0.3576054245473618,
      "hidden_pass_candidates_mean": 2.2125,
      "incremental_forward_tokens": 165846,
      "key": "default_k32",
      "name": "Default K32",
      "parse_success_mean": 9.7,
      "recovered": [
        22,
        35,
        36,
        42,
        67,
        70,
        81,
        87
      ],
      "remaining": [
        15,
        16,
        26,
        31,
        34,
        39,
        43,
        44,
        48,
        55,
        59,
        60,
        73,
        77,
        83,
        84
      ],
      "total_forward_tokens": 235491,
      "visible_repair_pass_count": 0,
      "zero_to_one": 8,
      "zero_to_one_rate": 0.3333333333333333
    },
    {
      "base_zero_subset": {
        "behavior_diversity_mean": 0.5393035299853061,
        "candidate_count_mean": 30.083333333333332,
        "functional_diversity_mean": 0.1256510899456186,
        "hidden_pass_candidates_mean": 1.4583333333333333,
        "parse_success_mean": 27.75
      },
      "behavior_diversity_rate_mean": 0.6847077256622585,
      "candidate_count_mean": 11.3875,
      "coverage": 0.825,
      "coverage_pct": 82.5,
      "false_repair_count": 0,
      "false_repair_rate": 0.0,
      "functional_diversity_rate_mean": 0.3585286603170189,
      "hidden_pass_candidates_mean": 2.3125,
      "incremental_forward_tokens": 173698,
      "key": "hot_k32",
      "name": "Hot K32",
      "parse_success_mean": 10.675,
      "recovered": [
        15,
        22,
        35,
        36,
        42,
        55,
        59,
        67,
        70,
        81
      ],
      "remaining": [
        16,
        26,
        31,
        34,
        39,
        43,
        44,
        48,
        60,
        73,
        77,
        83,
        84,
        87
      ],
      "total_forward_tokens": 243343,
      "visible_repair_pass_count": 0,
      "zero_to_one": 10,
      "zero_to_one_rate": 0.4166666666666667
    },
    {
      "base_zero_subset": {
        "behavior_diversity_mean": 0.5387479578533875,
        "candidate_count_mean": 28.916666666666668,
        "functional_diversity_mean": 0.13831285973944482,
        "hidden_pass_candidates_mean": 1.1666666666666667,
        "parse_success_mean": 26.291666666666668
      },
      "behavior_diversity_rate_mean": 0.6845410540226828,
      "candidate_count_mean": 11.0375,
      "coverage": 0.8125,
      "coverage_pct": 81.25,
      "false_repair_count": 0,
      "false_repair_rate": 0.0,
      "functional_diversity_rate_mean": 0.36232719125516677,
      "hidden_pass_candidates_mean": 2.225,
      "incremental_forward_tokens": 173623,
      "key": "diverse_k32",
      "name": "Diverse K32",
      "parse_success_mean": 10.2375,
      "recovered": [
        22,
        34,
        35,
        36,
        42,
        59,
        67,
        81,
        83
      ],
      "remaining": [
        15,
        16,
        26,
        31,
        39,
        43,
        44,
        48,
        55,
        60,
        70,
        73,
        77,
        84,
        87
      ],
      "total_forward_tokens": 243268,
      "visible_repair_pass_count": 0,
      "zero_to_one": 9,
      "zero_to_one_rate": 0.375
    },
    {
      "base_zero_subset": {
        "behavior_diversity_mean": 0.37692087292450777,
        "candidate_count_mean": 72.41666666666667,
        "functional_diversity_mean": 0.06394193005096116,
        "hidden_pass_candidates_mean": 3.3333333333333335,
        "parse_success_mean": 66.25
      },
      "behavior_diversity_rate_mean": 0.635992928544019,
      "candidate_count_mean": 24.0875,
      "coverage": 0.8625,
      "coverage_pct": 86.25,
      "false_repair_count": 0,
      "false_repair_rate": 0.0,
      "functional_diversity_rate_mean": 0.34001591234862166,
      "hidden_pass_candidates_mean": 2.875,
      "incremental_forward_tokens": 582812,
      "key": "union_k32",
      "name": "Union K32",
      "parse_success_mean": 22.225,
      "recovered": [
        15,
        22,
        34,
        35,
        36,
        42,
        55,
        59,
        67,
        70,
        81,
        83,
        87
      ],
      "remaining": [
        16,
        26,
        31,
        39,
        43,
        44,
        48,
        60,
        73,
        77,
        84
      ],
      "total_forward_tokens": 582812,
      "visible_repair_pass_count": 0,
      "zero_to_one": 13,
      "zero_to_one_rate": 0.5416666666666666
    },
    {
      "base_zero_subset": {
        "behavior_diversity_mean": 0.377304387636814,
        "candidate_count_mean": 73.0,
        "functional_diversity_mean": 0.06347874509547898,
        "hidden_pass_candidates_mean": 3.3333333333333335,
        "parse_success_mean": 66.83333333333333
      },
      "behavior_diversity_rate_mean": 0.6361079829577109,
      "candidate_count_mean": 24.2625,
      "coverage": 0.8625,
      "coverage_pct": 86.25,
      "false_repair_count": 2,
      "false_repair_rate": 1.0,
      "functional_diversity_rate_mean": 0.339876956861977,
      "hidden_pass_candidates_mean": 2.875,
      "incremental_forward_tokens": 12774,
      "key": "union_repair",
      "name": "Union+Repair",
      "parse_success_mean": 22.4,
      "recovered": [
        15,
        22,
        34,
        35,
        36,
        42,
        55,
        59,
        67,
        70,
        81,
        83,
        87
      ],
      "remaining": [
        16,
        26,
        31,
        39,
        43,
        44,
        48,
        60,
        73,
        77,
        84
      ],
      "total_forward_tokens": 595586,
      "visible_repair_pass_count": 2,
      "zero_to_one": 13,
      "zero_to_one_rate": 0.5416666666666666
    },
    {
      "base_zero_subset": {
        "behavior_diversity_mean": 0.3549131802370626,
        "candidate_count_mean": 89.83333333333333,
        "functional_diversity_mean": 0.05666196824427489,
        "hidden_pass_candidates_mean": 3.4583333333333335,
        "parse_success_mean": 81.95833333333333
      },
      "behavior_diversity_rate_mean": 0.6293906207377854,
      "candidate_count_mean": 29.3125,
      "coverage": 0.8875,
      "coverage_pct": 88.75,
      "false_repair_count": 0,
      "false_repair_rate": 0.0,
      "functional_diversity_rate_mean": 0.3378319238066158,
      "hidden_pass_candidates_mean": 2.9125,
      "incremental_forward_tokens": 115971,
      "key": "union_k128",
      "name": "Union K128",
      "parse_success_mean": 26.9375,
      "recovered": [
        15,
        22,
        34,
        35,
        36,
        42,
        55,
        59,
        67,
        70,
        73,
        81,
        83,
        84,
        87
      ],
      "remaining": [
        16,
        26,
        31,
        39,
        43,
        44,
        48,
        60,
        77
      ],
      "total_forward_tokens": 698783,
      "visible_repair_pass_count": 0,
      "zero_to_one": 15,
      "zero_to_one_rate": 0.625
    }
  ],
  "base_zero_task_ids": [
    15,
    16,
    22,
    26,
    31,
    34,
    35,
    36,
    39,
    42,
    43,
    44,
    48,
    55,
    59,
    60,
    67,
    70,
    73,
    77,
    81,
    83,
    84,
    87
  ],
  "experiment": "qwen35_4b_diversity_keyed_coverage_gate",
  "main_result": {
    "base_coverage": 0.7,
    "best_single_k32": "Hot K32",
    "best_single_k32_recovered": 10,
    "final_coverage": 0.8875,
    "final_recovered": 15,
    "union_k32_recovered": 13
  }
}