{
  "status": "terminal_verified_pending_independent_review_and_matched_evaluation",
  "stopping_reason": "total_deadline_soft_stop",
  "step": 75558,
  "unique_games": 2417856,
  "prediction_target_exposures": 551410959,
  "unique_corpus_train_targets": 943591445,
  "fraction_of_unique_target_corpus": 0.5843746908917767,
  "all_in_wall_seconds": 14284.400967359543,
  "initial_validation": {
    "nll": 5.029582231578336,
    "valid_targets": 9615601,
    "games": 42223,
    "phase_accuracy": {
      "source": 0.0005734652909349239,
      "destination": 0.0022277403023258876,
      "promotion": 1.0341177377515021e-05,
      "eos": 0.0
    },
    "phase_counts": {
      "source": 3191126,
      "destination": 3191126,
      "promotion": 3191126,
      "eos": 42223
    },
    "teacher_forced_triple_legality": 0.0,
    "raw_reserved_probability": 0.47164591255572563,
    "raw_illegal_token_probability": 0.9604235615133582,
    "legality_games": 128,
    "legality_targets": 29024,
    "legality_subset": "N evenly spaced game indices across immutable validation shard order; N=None means all",
    "protocol": "raw128 teacher-forced next-token logits; triple destination/promotion conditioned on gold prefix; not free-running",
    "seconds": 89.40276710100079
  },
  "validation_history": [
    {
      "epoch": 1,
      "complete_epoch": false,
      "step": 8022,
      "validation": {
        "nll": 0.8020281492112645,
        "valid_targets": 9615601,
        "games": 42223,
        "phase_accuracy": {
          "source": 0.49130369656353273,
          "destination": 0.6657502712208794,
          "promotion": 0.9994989229507076,
          "eos": 0.006299883949506193
        },
        "phase_counts": {
          "source": 3191126,
          "destination": 3191126,
          "promotion": 3191126,
          "eos": 42223
        },
        "teacher_forced_triple_legality": 0.552014119601329,
        "raw_reserved_probability": 6.802023497287675e-07,
        "raw_illegal_token_probability": 0.04695887935206488,
        "legality_games": 128,
        "legality_targets": 29024,
        "legality_subset": "N evenly spaced game indices across immutable validation shard order; N=None means all",
        "protocol": "raw128 teacher-forced next-token logits; triple destination/promotion conditioned on gold prefix; not free-running",
        "seconds": 90.02728238199961
      },
      "train_nll": 1.064056783467781,
      "useful_targets_per_second": 43851.233557608386,
      "padded_tokens_per_second": 43969.52108657463,
      "padding_fraction": 0.00269021645092149
    },
    {
      "epoch": 1,
      "complete_epoch": false,
      "step": 16044,
      "validation": {
        "nll": 0.7062413286860573,
        "valid_targets": 9615601,
        "games": 42223,
        "phase_accuracy": {
          "source": 0.5273897050758886,
          "destination": 0.715882732302015,
          "promotion": 0.9997439775176536,
          "eos": 0.03697037159841792
        },
        "phase_counts": {
          "source": 3191126,
          "destination": 3191126,
          "promotion": 3191126,
          "eos": 42223
        },
        "teacher_forced_triple_legality": 0.5880398671096345,
        "raw_reserved_probability": 8.015000277709161e-07,
        "raw_illegal_token_probability": 0.028326130378583142,
        "legality_games": 128,
        "legality_targets": 29024,
        "legality_subset": "N evenly spaced game indices across immutable validation shard order; N=None means all",
        "protocol": "raw128 teacher-forced next-token logits; triple destination/promotion conditioned on gold prefix; not free-running",
        "seconds": 89.2276000210004
      },
      "train_nll": 0.7479133022938583,
      "useful_targets_per_second": 43956.94552958297,
      "padded_tokens_per_second": 44067.373537030355,
      "padding_fraction": 0.0025058903806598476
    },
    {
      "epoch": 1,
      "complete_epoch": false,
      "step": 24066,
      "validation": {
        "nll": 0.6585811909788196,
        "valid_targets": 9615601,
        "games": 42223,
        "phase_accuracy": {
          "source": 0.5468878383366874,
          "destination": 0.7412809773102034,
          "promotion": 0.9999125700458082,
          "eos": 0.05688842573952585
        },
        "phase_counts": {
          "source": 3191126,
          "destination": 3191126,
          "promotion": 3191126,
          "eos": 42223
        },
        "teacher_forced_triple_legality": 0.6133720930232558,
        "raw_reserved_probability": 7.360765771569476e-07,
        "raw_illegal_token_probability": 0.019002528772384942,
        "legality_games": 128,
        "legality_targets": 29024,
        "legality_subset": "N evenly spaced game indices across immutable validation shard order; N=None means all",
        "protocol": "raw128 teacher-forced next-token logits; triple destination/promotion conditioned on gold prefix; not free-running",
        "seconds": 89.39242702399861
      },
      "train_nll": 0.681000865322067,
      "useful_targets_per_second": 43882.28545127481,
      "padded_tokens_per_second": 43998.63195933931,
      "padding_fraction": 0.0026443210364365255
    },
    {
      "epoch": 1,
      "complete_epoch": false,
      "step": 32088,
      "validation": {
        "nll": 0.6230701939552107,
        "valid_targets": 9615601,
        "games": 42223,
        "phase_accuracy": {
          "source": 0.5619141958042396,
          "destination": 0.7601081875175095,
          "promotion": 0.9999128834148198,
          "eos": 0.07919854107950643
        },
        "phase_counts": {
          "source": 3191126,
          "destination": 3191126,
          "promotion": 3191126,
          "eos": 42223
        },
        "teacher_forced_triple_legality": 0.6287375415282392,
        "raw_reserved_probability": 1.1706191127974988e-06,
        "raw_illegal_token_probability": 0.013474374564189399,
        "legality_games": 128,
        "legality_targets": 29024,
        "legality_subset": "N evenly spaced game indices across immutable validation shard order; N=None means all",
        "protocol": "raw128 teacher-forced next-token logits; triple destination/promotion conditioned on gold prefix; not free-running",
        "seconds": 89.22227741499955
      },
      "train_nll": 0.6396628259413834,
      "useful_targets_per_second": 43677.9978700924,
      "padded_tokens_per_second": 43792.34585464869,
      "padding_fraction": 0.002611140881464191
    },
    {
      "epoch": 1,
      "complete_epoch": false,
      "step": 40110,
      "validation": {
        "nll": 0.5984513133298186,
        "valid_targets": 9615601,
        "games": 42223,
        "phase_accuracy": {
          "source": 0.5726608726825578,
          "destination": 0.7732687458909488,
          "promotion": 0.99992165774714,
          "eos": 0.11848992255405821
        },
        "phase_counts": {
          "source": 3191126,
          "destination": 3191126,
          "promotion": 3191126,
          "eos": 42223
        },
        "teacher_forced_triple_legality": 0.637250830564784,
        "raw_reserved_probability": 1.2146030691359047e-06,
        "raw_illegal_token_probability": 0.010478415393755354,
        "legality_games": 128,
        "legality_targets": 29024,
        "legality_subset": "N evenly spaced game indices across immutable validation shard order; N=None means all",
        "protocol": "raw128 teacher-forced next-token logits; triple destination/promotion conditioned on gold prefix; not free-running",
        "seconds": 89.56453253000291
      },
      "train_nll": 0.6096019479880465,
      "useful_targets_per_second": 44085.821135475,
      "padded_tokens_per_second": 44203.19083324089,
      "padding_fraction": 0.002655231343109965
    },
    {
      "epoch": 1,
      "complete_epoch": false,
      "step": 48132,
      "validation": {
        "nll": 0.5853722610531451,
        "valid_targets": 9615601,
        "games": 42223,
        "phase_accuracy": {
          "source": 0.5789667972997619,
          "destination": 0.7802756769867438,
          "promotion": 0.9999219711161514,
          "eos": 0.11867939274802833
        },
        "phase_counts": {
          "source": 3191126,
          "destination": 3191126,
          "promotion": 3191126,
          "eos": 42223
        },
        "teacher_forced_triple_legality": 0.6434800664451827,
        "raw_reserved_probability": 1.4037503132838264e-06,
        "raw_illegal_token_probability": 0.008921839463868528,
        "legality_games": 128,
        "legality_targets": 29024,
        "legality_subset": "N evenly spaced game indices across immutable validation shard order; N=None means all",
        "protocol": "raw128 teacher-forced next-token logits; triple destination/promotion conditioned on gold prefix; not free-running",
        "seconds": 88.78254190799998
      },
      "train_nll": 0.5900576727012877,
      "useful_targets_per_second": 44042.16542362308,
      "padded_tokens_per_second": 44158.326903673595,
      "padding_fraction": 0.0026305679629553547
    },
    {
      "epoch": 1,
      "complete_epoch": false,
      "step": 56154,
      "validation": {
        "nll": 0.5806068211344487,
        "valid_targets": 9615601,
        "games": 42223,
        "phase_accuracy": {
          "source": 0.5815580456553581,
          "destination": 0.7826889317438421,
          "promotion": 0.9999219711161514,
          "eos": 0.12625820050683276
        },
        "phase_counts": {
          "source": 3191126,
          "destination": 3191126,
          "promotion": 3191126,
          "eos": 42223
        },
        "teacher_forced_triple_legality": 0.645141196013289,
        "raw_reserved_probability": 1.2305505497339278e-06,
        "raw_illegal_token_probability": 0.008621482281061693,
        "legality_games": 128,
        "legality_targets": 29024,
        "legality_subset": "N evenly spaced game indices across immutable validation shard order; N=None means all",
        "protocol": "raw128 teacher-forced next-token logits; triple destination/promotion conditioned on gold prefix; not free-running",
        "seconds": 87.16885058499975
      },
      "train_nll": 0.5827513558489217,
      "useful_targets_per_second": 44833.23142495002,
      "padded_tokens_per_second": 44950.27749682339,
      "padding_fraction": 0.0026039009855201023
    },
    {
      "epoch": 1,
      "complete_epoch": false,
      "step": 64176,
      "validation": {
        "nll": 0.5771529146043254,
        "valid_targets": 9615601,
        "games": 42223,
        "phase_accuracy": {
          "source": 0.5831411859011522,
          "destination": 0.7842720719896362,
          "promotion": 0.9999219711161514,
          "eos": 0.1320133576486749
        },
        "phase_counts": {
          "source": 3191126,
          "destination": 3191126,
          "promotion": 3191126,
          "eos": 42223
        },
        "teacher_forced_triple_legality": 0.645764119601329,
        "raw_reserved_probability": 1.6971123869284854e-06,
        "raw_illegal_token_probability": 0.008022273886824883,
        "legality_games": 128,
        "legality_targets": 29024,
        "legality_subset": "N evenly spaced game indices across immutable validation shard order; N=None means all",
        "protocol": "raw128 teacher-forced next-token logits; triple destination/promotion conditioned on gold prefix; not free-running",
        "seconds": 87.28590715500104
      },
      "train_nll": 0.5785217669855705,
      "useful_targets_per_second": 45314.38481190093,
      "padded_tokens_per_second": 45429.73905951081,
      "padding_fraction": 0.0025391791808175856
    },
    {
      "epoch": 1,
      "complete_epoch": false,
      "step": 72198,
      "validation": {
        "nll": 0.5739030495971976,
        "valid_targets": 9615601,
        "games": 42223,
        "phase_accuracy": {
          "source": 0.5851229315294978,
          "destination": 0.7861008935403992,
          "promotion": 0.9999219711161514,
          "eos": 0.12218459133647538
        },
        "phase_counts": {
          "source": 3191126,
          "destination": 3191126,
          "promotion": 3191126,
          "eos": 42223
        },
        "teacher_forced_triple_legality": 0.6450373754152824,
        "raw_reserved_probability": 1.2542378441883736e-06,
        "raw_illegal_token_probability": 0.007861457974170486,
        "legality_games": 128,
        "legality_targets": 29024,
        "legality_subset": "N evenly spaced game indices across immutable validation shard order; N=None means all",
        "protocol": "raw128 teacher-forced next-token logits; triple destination/promotion conditioned on gold prefix; not free-running",
        "seconds": 86.81734904599944
      },
      "train_nll": 0.574783663082018,
      "useful_targets_per_second": 45393.52858503474,
      "padded_tokens_per_second": 45510.97691784213,
      "padding_fraction": 0.0025806594531997407
    }
  ],
  "best_nll": 0.5739030495971976,
  "checkpoints": {
    "best": {
      "sha256": "857bc93bfffb287dc67e2845a14c214a4691bc1a008dae1fa4d044e5fccc87d2",
      "step": 72198,
      "bytes": 1031498263,
      "targets_at_checkpoint": 526834590
    },
    "last": {
      "sha256": "d000224f697a891687a74a8cce871d60024af3c3e536ec220fd21567eb350dc9",
      "step": 75558,
      "bytes": 1031498263,
      "targets_at_checkpoint": 551410959
    }
  },
  "test_evaluated": false
}