{ "rules": "hex-a-hop-basic-v1", "observation": "hex-basic-start-centered-safe-green-player-v1", "scope": "generated NORMAL/COLLAPSABLE boards, 4-8 green tiles; not the full original campaign", "config": { "seed": 47, "split_seed": 20261015, "train_boards": 256, "validation_boards": 64, "test_boards": 64, "max_steps": 48, "bc_epochs": 100, "ppo_steps": 65536, "cpu_threads": 1, "wall_seconds": 300 }, "bc_positions": 1825, "ppo_actual_steps": 65536, "bc_ppo_actual_steps": 65536, "wall_seconds": 17.79851895896718, "versions": { "python": "3.12.13", "torch": "2.14.1" }, "hashes": { "teacher.json": "d50211429c9c052071251adb8b0f73a74e073b89b24992d98146419490e227ef", "bc.zip": "25eb6c37807c6b76d345ea50effcd8d7bc65ae63f905397761c6d7e71781c231", "splits.json": "9e389be15b60712dd858b075be281c70d3d5f5d1786e4d21c44c68bd1ca4ff9f", "config.json": "34c77a018976991cda87b0885f3ad225955f1fe57707a3e4c31947541e1e8972", "ppo.zip": "c149a54d2c924834a510d399e3e32a0a2f764dcc9a765551d0149fc7f301cd20", "bc-ppo.zip": "c866468a4c0e7ca1cc623b4ed5ca1b812e8e49c310b8c5bb9518fe193536bb50" }, "independent_training_seeds": 1, "qualification": false, "results": { "train": { "random-six": { "wins": 40, "episodes": 256, "win_rate": 0.15625, "mean_moves": 44.8359375, "invalid_fraction": 0.8885694371841785 }, "random-legal": { "wins": 52, "episodes": 256, "win_rate": 0.203125, "mean_moves": 39.62890625, "invalid_fraction": 0.8534253326761951 }, "bc": { "wins": 185, "episodes": 256, "win_rate": 0.72265625, "mean_moves": 18.34765625, "invalid_fraction": 0.6536086863955717 }, "ppo": { "wins": 0, "episodes": 256, "win_rate": 0.0, "mean_moves": 48.0, "invalid_fraction": 0.9671223958333334 }, "bc-ppo": { "wins": 118, "episodes": 256, "win_rate": 0.4609375, "mean_moves": 29.06640625, "invalid_fraction": 0.8215293643327509 }, "search": { "wins": 256, "episodes": 256, "max_nodes": 50000, "mean_expanded": 68.9609375, "statuses": [ "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved" ] } }, "validation": { "random-six": { "wins": 8, "episodes": 64, "win_rate": 0.125, "mean_moves": 45.890625, "invalid_fraction": 0.8985359210078311 }, "random-legal": { "wins": 8, "episodes": 64, "win_rate": 0.125, "mean_moves": 42.9375, "invalid_fraction": 0.8704512372634643 }, "bc": { "wins": 2, "episodes": 64, "win_rate": 0.03125, "mean_moves": 46.65625, "invalid_fraction": 0.9457468184862693 }, "ppo": { "wins": 0, "episodes": 64, "win_rate": 0.0, "mean_moves": 48.0, "invalid_fraction": 0.9674479166666666 }, "bc-ppo": { "wins": 3, "episodes": 64, "win_rate": 0.046875, "mean_moves": 46.015625, "invalid_fraction": 0.9477079796264856 }, "search": { "wins": 64, "episodes": 64, "max_nodes": 50000, "mean_expanded": 46.25, "statuses": [ "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved" ] } }, "test": { "random-six": { "wins": 10, "episodes": 64, "win_rate": 0.15625, "mean_moves": 45.140625, "invalid_fraction": 0.8843890619591555 }, "random-legal": { "wins": 11, "episodes": 64, "win_rate": 0.171875, "mean_moves": 41.140625, "invalid_fraction": 0.8458032662362325 }, "bc": { "wins": 2, "episodes": 64, "win_rate": 0.03125, "mean_moves": 46.703125, "invalid_fraction": 0.9424556707929074 }, "ppo": { "wins": 0, "episodes": 64, "win_rate": 0.0, "mean_moves": 48.0, "invalid_fraction": 0.9690755208333334 }, "bc-ppo": { "wins": 1, "episodes": 64, "win_rate": 0.015625, "mean_moves": 47.34375, "invalid_fraction": 0.9491749174917492 }, "search": { "wins": 64, "episodes": 64, "max_nodes": 50000, "mean_expanded": 72.515625, "statuses": [ "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved", "solved" ] } } }, "original_basic_search": [ { "name": "Levels\\0_green\\asymmetrix.lev", "green": 47, "status": "budget", "expanded": 50000, "solution_moves": null }, { "name": "Levels\\0_green\\hive.lev", "green": 24, "status": "solved", "expanded": 5802, "solution_moves": 24 }, { "name": "Levels\\0_green\\there and back.lev", "green": 21, "status": "solved", "expanded": 14289, "solution_moves": 32 }, { "name": "Levels\\0_green\\triangular.lev", "green": 32, "status": "solved", "expanded": 4217, "solution_moves": 33 } ], "source_verification": { "passed": true, "boards": 104, "trajectories": 207, "transitions": 14719, "upstream_sha256": "7c53916830db8f15fff87b4873acb83693d63b66e74cb36bcd39507bce1d56c0", "scope": "Compare Python transitions with a source-extracted C++ basic-rule oracle.\n\nThis is NOT full native/browser-engine parity. The oracle compiles the pinned\nupstream position operator, water/collision gate, green departure branch and\nCheckFinished method; rendering/sound and unsupported tile mechanics are omitted.\nExtracted upstream code is GPL-2.0-or-later, Tom Beaumont, 2005-2007.\n", "oracle_cpp_sha256": "e5ea50602ebc17baa2719c2b454c835c3ed47db068de81f18ed0f0c05defda31" }, "checkpoint_reload_replay": { "passed": true, "episodes": 1152, "comparison": "every action and outcome across train, validation and test for all three checkpoints" }, "results_sha256": "df934a15914541f09e3c3872f11dae87e973e7b4ab924618e86488aa9c43270b", "code_sha256": { "verify_source.py": "eae551c208bee3a391e5e369a6cd558820e7fac989e7602d026fac5c63f78306", "run.py": "2bbc1e0fb6314d28432835c99ea0c60ddc611b206691262f015ba1d40fc36817", "env.py": "8cbba180fbb920bf0e63045ccafbcab830f107a4d8d51dcddd95f7aac12dd30c", "engine.py": "83ac1d389dbe6044dbf14123a294f177b1224b9c2b53bd56b7a92e1e7564ae85", "boards.py": "19856af199b1c840674249e9030a1f7be58d3a31319e32a1d61d779492b1df6c" } }