{
  "generated_utc": "2026-10-04T16:37:05.215012+00:00",
  "evaluation_policy": "greedy argmax Q; current and next fruit; no future-seed lookahead",
  "evaluation_suites": {
    "448x720": "seeds 0-11 (12 per checkpoint)",
    "550x720": "seeds 0-3 (4 per board per checkpoint)"
  },
  "sampling_note": "Fixed development suites reused during training; means are not held-out estimates.",
  "curve_note": "All evaluation records retained; lines break when the environment-step counter rolls back after a restart.",
  "censoring_note": "Triangle markers denote cells containing capped episodes; those scores are lower bounds.",
  "continuation_note": "w6c runs preserve weights and environment-step counters, restart learning-rate schedules and gradient counters.",
  "loss_note": "Median of finite positive logged losses per 10-million environment-step bin; no gameplay claim follows from loss.",
  "sources": [
    {
      "path": "dqn_runs/a100_t1_2/wave6_20260930/w6s_h720/eval.jsonl",
      "sha256": "6a251f6ef530d654218913489b08a6287b07148da17fa1cddb52bd4b79e1b32a"
    },
    {
      "path": "dqn_runs/a100_t1_2/wave6_20260930/w6s_h720/metrics.jsonl",
      "sha256": "52e21e07032c7751fb2e7bd0d4fb0ba0e1b655036e138af55a68ff229308d016"
    },
    {
      "path": "dqn_runs/a100_t1_2/wave6_20260930/w6d_h720/eval.jsonl",
      "sha256": "7328028317c35279640d3e46ce42eff7adf116291997f047b5eb2690e14a63fd"
    },
    {
      "path": "dqn_runs/a100_t1_2/wave6_20260930/w6d_h720/metrics.jsonl",
      "sha256": "7f8bc6edeac132cf0eec482235ca4862d71790ed261cf30314dfb967a9ed5e1d"
    },
    {
      "path": "dqn_runs/a100_t1_2/wave6_20260930/w6c_h720/eval.jsonl",
      "sha256": "91e25376a7960a447f2c3dd6e49481e445e79dca6e42d6022af2582403a83f10"
    },
    {
      "path": "dqn_runs/a100_t1_2/wave6_20260930/w6c_h720/metrics.jsonl",
      "sha256": "124547168dd9486cd6a864b158a38e5f3d3e79b177ce452cbb8a4484e0e751bf"
    },
    {
      "path": "dqn_runs/a100_perm/wave6_20260930/w6s_m720/eval.jsonl",
      "sha256": "5595b553e671508aca10ce97a3343c2054b50b83a26a8dd39f2fd4440d90c7f3"
    },
    {
      "path": "dqn_runs/a100_perm/wave6_20260930/w6s_m720/metrics.jsonl",
      "sha256": "3fc317df5f4a8ce44278ea79899d90dc399c03d8850dee00b8fe957cf4ab2809"
    },
    {
      "path": "dqn_runs/a100_perm/wave6_20260930/w6c_m720/eval.jsonl",
      "sha256": "b47fba394a5fcb1173e4181bf7d0eb0bab017ce863e47716c59f68f01eb0bac9"
    },
    {
      "path": "dqn_runs/a100_perm/wave6_20260930/w6c_m720/metrics.jsonl",
      "sha256": "b7550de7ba65bba7fa32b3d5b37d74f3bef0aa0cef95e489c97ad2d1d195c013"
    },
    {
      "path": "dqn_runs/a100_t1_3/wave6_20260930/w6s_var/eval.jsonl",
      "sha256": "c0bb4b06a6e2d3e9285eec60960da29ac0d37157c76d73bee6de0d4a094c70f3"
    },
    {
      "path": "dqn_runs/a100_t1_3/wave6_20260930/w6s_var/metrics.jsonl",
      "sha256": "0a8d168fe09df789be7bc4b2ac8ecc33e6016cf0d54bfdc16a1cc61fed20a86e"
    },
    {
      "path": "dqn_runs/a100_t1_3/wave6_20260930/w6c_var/eval.jsonl",
      "sha256": "ecd150f2c25ccde79f31f97e190a03bed5110155f29f2f8956ad4de67a99519b"
    },
    {
      "path": "dqn_runs/a100_t1_3/wave6_20260930/w6c_var/metrics.jsonl",
      "sha256": "1324ad4720cce7339e1cddd698ddac9ba036eecfc63c6294d860579c156b1a8a"
    },
    {
      "path": "dqn_runs/a100_t1/wave6_20260930/w6s_mvar/eval.jsonl",
      "sha256": "895b19df590b781ef6726daa9235bf7f790d4ce61249a28cf10280a64cf8d720"
    },
    {
      "path": "dqn_runs/a100_t1/wave6_20260930/w6s_mvar/metrics.jsonl",
      "sha256": "a10127d2212a9fe1bacf8e5fcd247f3e49a9449c891f4991305abd01c559fd45"
    },
    {
      "path": "dqn_runs/a100_t1/wave6_20260930/w6c_mvar/eval.jsonl",
      "sha256": "6b4a7c40a3964a6734039976056baacf35e77b7f87af890eb1990f1e5db2665e"
    },
    {
      "path": "dqn_runs/a100_t1/wave6_20260930/w6c_mvar/metrics.jsonl",
      "sha256": "7134588870475b23fb2bfcbfefc5c1e8706bb901cc40cd49cab1287070d8fe85"
    }
  ],
  "summary": {
    "448x720/Set Transformer": {
      "best_uncensored_evaluation_mean": {
        "run": "w6s_h720",
        "grad_steps": 12712,
        "mean": 6898.75,
        "median": 4973.5,
        "max": 18828.0,
        "moves_mean": 623.9166666666666,
        "maxfruit_max": 10,
        "censored": 0,
        "n": 12
      },
      "final_evaluation": {
        "run": "w6c_h720",
        "grad_steps": 14437,
        "mean": 4348.75,
        "median": 3969.0,
        "max": 7982.0,
        "moves_mean": 408.1666666666667,
        "maxfruit_max": 10,
        "censored": 0,
        "n": 12
      }
    },
    "448x720/MLP": {
      "best_uncensored_evaluation_mean": {
        "run": "w6c_m720",
        "grad_steps": 3357,
        "mean": 2932.5833333333335,
        "median": 2593.5,
        "max": 6578.0,
        "moves_mean": 295.25,
        "maxfruit_max": 10,
        "censored": 0,
        "n": 12
      },
      "final_evaluation": {
        "run": "w6c_m720",
        "grad_steps": 74789,
        "mean": 2520.4166666666665,
        "median": 2426.0,
        "max": 3825.0,
        "moves_mean": 258.4166666666667,
        "maxfruit_max": 10,
        "censored": 0,
        "n": 12
      }
    },
    "550x720/Set Transformer": {
      "best_uncensored_evaluation_mean": {
        "run": "w6s_var",
        "grad_steps": 14893,
        "mean": 14422.25,
        "median": 13566.5,
        "max": 25291.0,
        "moves_mean": 1261.0,
        "maxfruit_max": 10,
        "censored": 0,
        "n": 4
      },
      "final_evaluation": {
        "run": "w6c_var",
        "grad_steps": 13529,
        "mean": 8033.0,
        "median": 7918.0,
        "max": 12065.0,
        "moves_mean": 730.25,
        "maxfruit_max": 10,
        "censored": 0,
        "n": 4
      }
    },
    "550x720/MLP": {
      "best_uncensored_evaluation_mean": {
        "run": "w6s_mvar",
        "grad_steps": 183477,
        "mean": 5690.5,
        "median": 5717.0,
        "max": 7740.0,
        "moves_mean": 540.75,
        "maxfruit_max": 10,
        "censored": 0,
        "n": 4
      },
      "final_evaluation": {
        "run": "w6c_mvar",
        "grad_steps": 66292,
        "mean": 3789.5,
        "median": 4217.5,
        "max": 4300.0,
        "moves_mean": 368.75,
        "maxfruit_max": 10,
        "censored": 0,
        "n": 4
      }
    }
  }
}
