{ "schema_version": "nanojev-sonic-supervision-summary-v1", "passed": true, "created_at": "2026-09-20T04:48:32.753964+00:00", "cases": { "path": "runs/sonic_unified_sft_v1/test_cases.jsonl", "sha256": "d00f5b6aeaa9f2aac3e4b5010b31df4e86f77674cb588e17bc3f7d853e4a1f36", "episodes": 548, "counts": { "test/maze": 10, "test/snake": 8, "ood/maze": 10, "ood/snake": 8, "test/basic": 128, "ood/basic": 128, "test/predict_position": 128, "ood/predict_position": 128 } }, "experiment": { "path": "runs/sonic_unified_sft_v1/experiment/experiment.json", "sha256": "cb4937280e24dc1f089431dbad29ee7185acea17daae1e9b470266fba0abda7e", "selection_sha256": "1b9dde19e02abfdafc74dfcb724b203952932dacac64950993f0e493c9670f59", "selection": { "selected": "hard_lr1e5", "best_new_arm": "hard_lr1e5", "checkpoint": "/data/rwang/nanojev_sonic_sft_20260920/experiment/hard_lr1e5", "weights_sha256": "f68c47d66998231b86b7e91b4ed5e82ae23acf104c8b7cd6d165c3ac7b7ffe1b", "dev_results": { "initial": { "weighted_success": 0.6963541666666666, "rates": { "maze/policy": 0.8, "snake/policy": 0.75, "shooting/policy/basic": 1.0, "shooting/policy/predict_position": 0.078125 }, "counts": { "maze/policy": { "successes": 8, "episodes": 10 }, "snake/policy": { "successes": 6, "episodes": 8 }, "shooting/policy/basic": { "successes": 64, "episodes": 64 }, "shooting/policy/predict_position": { "successes": 5, "episodes": 64 } }, "by_split": { "dev": { "maze/policy": { "successes": 8, "episodes": 10, "success_rate": 0.8 }, "snake/policy": { "successes": 6, "episodes": 8, "success_rate": 0.75 }, "shooting/policy/basic": { "successes": 64, "episodes": 64, "success_rate": 1.0 }, "shooting/policy/predict_position": { "successes": 5, "episodes": 64, "success_rate": 0.078125 } } }, "episodes_sha256": "4c96065e61b0990809550b06f5e79a29d9887496093b7b0c3a547e00b8f5a386" }, "soft_lr1e5": { "weighted_success": 0.7484375, "rates": { "maze/policy": 0.8, "snake/policy": 0.875, "shooting/policy/basic": 1.0, "shooting/policy/predict_position": 0.140625 }, "counts": { "maze/policy": { "successes": 8, "episodes": 10 }, "snake/policy": { "successes": 7, "episodes": 8 }, "shooting/policy/basic": { "successes": 64, "episodes": 64 }, "shooting/policy/predict_position": { "successes": 9, "episodes": 64 } }, "by_split": { "dev": { "maze/policy": { "successes": 8, "episodes": 10, "success_rate": 0.8 }, "snake/policy": { "successes": 7, "episodes": 8, "success_rate": 0.875 }, "shooting/policy/basic": { "successes": 64, "episodes": 64, "success_rate": 1.0 }, "shooting/policy/predict_position": { "successes": 9, "episodes": 64, "success_rate": 0.140625 } } }, "episodes_sha256": "b8904b28a032887f4847e26053d2dd06450cdcd8744a77c4bebbc465215ba124" }, "hard_lr2e5": { "weighted_success": 0.6734374999999999, "rates": { "maze/policy": 0.7, "snake/policy": 0.75, "shooting/policy/basic": 1.0, "shooting/policy/predict_position": 0.140625 }, "counts": { "maze/policy": { "successes": 7, "episodes": 10 }, "snake/policy": { "successes": 6, "episodes": 8 }, "shooting/policy/basic": { "successes": 64, "episodes": 64 }, "shooting/policy/predict_position": { "successes": 9, "episodes": 64 } }, "by_split": { "dev": { "maze/policy": { "successes": 7, "episodes": 10, "success_rate": 0.7 }, "snake/policy": { "successes": 6, "episodes": 8, "success_rate": 0.75 }, "shooting/policy/basic": { "successes": 64, "episodes": 64, "success_rate": 1.0 }, "shooting/policy/predict_position": { "successes": 9, "episodes": 64, "success_rate": 0.140625 } } }, "episodes_sha256": "bec27b28424b25fce356d14d5ea318b91f105e3812bc96ba0074696f1b39b2ab" }, "hard_lr1e5": { "weighted_success": 0.803125, "rates": { "maze/policy": 0.8, "snake/policy": 1.0, "shooting/policy/basic": 1.0, "shooting/policy/predict_position": 0.21875 }, "counts": { "maze/policy": { "successes": 8, "episodes": 10 }, "snake/policy": { "successes": 8, "episodes": 8 }, "shooting/policy/basic": { "successes": 64, "episodes": 64 }, "shooting/policy/predict_position": { "successes": 14, "episodes": 64 } }, "by_split": { "dev": { "maze/policy": { "successes": 8, "episodes": 10, "success_rate": 0.8 }, "snake/policy": { "successes": 8, "episodes": 8, "success_rate": 1.0 }, "shooting/policy/basic": { "successes": 64, "episodes": 64, "success_rate": 1.0 }, "shooting/policy/predict_position": { "successes": 14, "episodes": 64, "success_rate": 0.21875 } } }, "episodes_sha256": "e4d7b320d0c57c5933821b16fdae69fbac172bdac67e31c4b4ffd0b08c912db0" }, "soft_lr2e5": { "weighted_success": 0.6682291666666667, "rates": { "maze/policy": 0.7, "snake/policy": 0.75, "shooting/policy/basic": 0.953125, "shooting/policy/predict_position": 0.15625 }, "counts": { "maze/policy": { "successes": 7, "episodes": 10 }, "snake/policy": { "successes": 6, "episodes": 8 }, "shooting/policy/basic": { "successes": 61, "episodes": 64 }, "shooting/policy/predict_position": { "successes": 10, "episodes": 64 } }, "by_split": { "dev": { "maze/policy": { "successes": 7, "episodes": 10, "success_rate": 0.7 }, "snake/policy": { "successes": 6, "episodes": 8, "success_rate": 0.75 }, "shooting/policy/basic": { "successes": 61, "episodes": 64, "success_rate": 0.953125 }, "shooting/policy/predict_position": { "successes": 10, "episodes": 64, "success_rate": 0.15625 } } }, "episodes_sha256": "af2aab6607e729af66fb65db78e3ee081361e41e433cf293359938b604f7a052" } }, "test_results_used": false, "selection_rule": "Highest weighted dev episode success including the unchanged initialization; ties prefer higher PP success, then Basic success, then initialization, then arm declaration order." }, "dev_selection_note": "Rule recomputed from recorded dev scores; no heldout outcomes used. This script does not rerun dev models." }, "primary_controller": { "controller": "greedy", "epsilon": 0.1, "sampling_seed": 17, "temperature": 1.0 }, "runs": { "initial": { "source": { "episodes_path": "runs/sonic_unified_sft_v1/experiment/initial_test.jsonl", "manifest_path": "runs/sonic_unified_sft_v1/experiment/initial_test.manifest.json", "episodes_sha256": "6845880e1e6c8ff6ed3aedd2be90762d24c0563d67a3ab29ff452c3572e7b795", "manifest_sha256": "82aa1efe44ae7a12b1b7bd429fe6e6140620ac5aeeef8fc1946b861f2337b8ef", "cases_sha256": "d00f5b6aeaa9f2aac3e4b5010b31df4e86f77674cb588e17bc3f7d853e4a1f36", "continuation_policy_id": "564c891a6a97e0acc1743800c9ee5358695854e052aff2c9957dc8a7600daa4a", "policy": { "engine": "checkpoint", "controller": "greedy", "epsilon": 0.1, "sampling_seed": 17, "temperature": 1.0, "source_sha256": { "unified_game_pipeline.py": "c176e06d7088bafbf0fbe4ca7e2be05eef6f1f5c3bf00b2d946caa21cce42313", "unified_grid_envs.py": "8ab8a9ff56f88866efc26bca41c16416e432d5ed8415cdcd33f622c5c7030663", "unified_doom_env.py": "fba56e3e97fe7bf0940962814308f63fcd2a788bf930e36f7f75b2aaa1889039", "scaled_maze.py": "4eef4a1052c913755ccccc54b9e7ee2cca5e6a1258ab748d59bfd45386d410bb", "snake_game.py": "adacc7a6013f0195e5a0952bc2a13ec14964325b52b56b2e69d780794edf549e", "evaluate_composed_maze.py": "b443469cedf81b95c154e33a7d90accd9bb7cd7423a40bfd7314be1f42fc9ae5", "predict_toy_decisions.py": "6eedc49aaf9a81763677e3cff4e87fc4e499bfc5decf0bb4678f452f578dd92b", "train_toy_decisions.py": "4f39babd5575e43d7acf3eca357329a93c60041f334bc2bd7424929bf9c0da86", "unified_jev_worker.mjs": "e844d42fcf6f1ac844c7525746cef3b8d87848682ccbaafe91e1c12294bc1dbf", "teachers.mjs": "6f8dd533e34a8f491530a26197081461e70977fce1857924ac5856c4817787ab" }, "tie_break": "lexicographic_first", "environment_contract": "finite_task_deadline_v1", "checkpoint_sha256": { "config.json": "e17d5ed4aef12e07f73d380080d395c512a004c72398ef6c7f1d8055edf51442", "best.safetensors": "38116340795de1c82369b7fe15819d92d79600a7b4dc7a3cd0d4390cb6782639", "tokenizer/tokenizer.json": "be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506", "backbone_config/config.json": "30c011854471509747858ac07ffb8a4dab2bc6c04035b184729b69500c557c54" } } }, "validation": { "matched_episodes": 548, "controller_transitions_checked": 12100, "physical_replay_performed_by_this_script": false }, "by_split_category": { "test/maze": { "n": 10, "successes": 4, "success_rate": 0.4, "wilson95": [ 0.16818032970623617, 0.6873262302663417 ], "mean_steps": 187.4, "mean_decisions": 152.3, "metrics": { "collisions": { "n": 10, "mean": 92.4, "min": 7.0, "max": 363.0 }, "decision_steps": { "n": 10, "mean": 187.4, "min": 21.0, "max": 724.0 }, "physical_steps": { "n": 10, "mean": 207.6, "min": 21.0, "max": 800.0 }, "remaining_steps": { "n": 10, "mean": 3.6, "min": 0.0, "max": 11.0 }, "reposition_steps": { "n": 10, "mean": 20.2, "min": 0.0, "max": 100.0 }, "visited_cells": { "n": 10, "mean": 96.0, "min": 14.0, "max": 362.0 } } }, "test/snake": { "n": 8, "successes": 6, "success_rate": 0.75, "wilson95": [ 0.40927543031016894, 0.928520787247891 ], "mean_steps": 19.25, "mean_decisions": 19.25, "metrics": { "body_length": { "n": 8, "mean": 4.75, "min": 3.0, "max": 5.0 }, "collisions": { "n": 8, "mean": 0.25, "min": 0.0, "max": 1.0 }, "decision_steps": { "n": 8, "mean": 19.25, "min": 4.0, "max": 43.0 }, "food_collected": { "n": 8, "mean": 1.75, "min": 0.0, "max": 2.0 }, "initial_score": { "n": 8, "mean": 0.0, "min": 0.0, "max": 0.0 }, "physical_steps": { "n": 8, "mean": 19.25, "min": 4.0, "max": 43.0 }, "remaining_steps": { "n": 8, "mean": 92.75, "min": 64.0, "max": 127.0 }, "score": { "n": 8, "mean": 1.75, "min": 0.0, "max": 2.0 }, "survival_steps": { "n": 8, "mean": 19.0, "min": 4.0, "max": 42.0 }, "target_food": { "n": 8, "mean": 2.25, "min": 2.0, "max": 3.0 } } }, "test/basic": { "n": 128, "successes": 128, "success_rate": 1.0, "wilson95": [ 0.9708630437264916, 1.0 ], "mean_steps": 6.2734375, "mean_decisions": 6.2734375, "metrics": { "ammo": { "n": 128, "mean": 48.8828125, "min": 47.0, "max": 49.0 }, "ammo_consumed": { "n": 128, "mean": 1.1171875, "min": 1.0, "max": 3.0 }, "damage_dealt": { "n": 128, "mean": 9.7265625, "min": 5.0, "max": 15.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 6.2734375, "min": 2.0, "max": 17.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 1.0, "min": 1.0, "max": 1.0 }, "kills": { "n": 128, "mean": 1.0, "min": 1.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": 78.1015625, "min": 24.0, "max": 96.0 }, "physical_ticks": { "n": 128, "mean": 22.3125, "min": 5.0, "max": 67.0 } } }, "test/predict_position": { "n": 128, "successes": 11, "success_rate": 0.0859375, "wilson95": [ 0.0486631313713862, 0.14734091054261297 ], "mean_steps": 26.8046875, "mean_decisions": 26.8046875, "metrics": { "ammo": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "ammo_consumed": { "n": 128, "mean": 1.0, "min": 1.0, "max": 1.0 }, "damage_dealt": { "n": 128, "mean": 16.2421875, "min": 0.0, "max": 208.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 26.8046875, "min": 10.0, "max": 71.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.0859375, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.0859375, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": -0.02023437500000007, "min": -0.2840000000000002, "max": 0.962 }, "physical_ticks": { "n": 128, "mean": 106.171875, "min": 38.0, "max": 284.0 } } }, "ood/maze": { "n": 10, "successes": 2, "success_rate": 0.2, "wilson95": [ 0.05668215145437522, 0.5098375284633582 ], "mean_steps": 265.2, "mean_decisions": 222.5, "metrics": { "collisions": { "n": 10, "mean": 125.1, "min": 71.0, "max": 361.0 }, "decision_steps": { "n": 10, "mean": 265.2, "min": 142.0, "max": 764.0 }, "physical_steps": { "n": 10, "mean": 286.6, "min": 147.0, "max": 800.0 }, "remaining_steps": { "n": 10, "mean": 1.4, "min": 0.0, "max": 13.0 }, "reposition_steps": { "n": 10, "mean": 21.4, "min": 0.0, "max": 114.0 }, "visited_cells": { "n": 10, "mean": 141.1, "min": 62.0, "max": 478.0 } } }, "ood/snake": { "n": 8, "successes": 4, "success_rate": 0.5, "wilson95": [ 0.2152160622138776, 0.7847839377861224 ], "mean_steps": 16.125, "mean_decisions": 16.125, "metrics": { "body_length": { "n": 8, "mean": 5.0, "min": 3.0, "max": 6.0 }, "collisions": { "n": 8, "mean": 0.5, "min": 0.0, "max": 1.0 }, "decision_steps": { "n": 8, "mean": 16.125, "min": 7.0, "max": 35.0 }, "food_collected": { "n": 8, "mean": 2.0, "min": 0.0, "max": 3.0 }, "initial_score": { "n": 8, "mean": 0.0, "min": 0.0, "max": 0.0 }, "physical_steps": { "n": 8, "mean": 16.125, "min": 7.0, "max": 35.0 }, "remaining_steps": { "n": 8, "mean": 143.875, "min": 125.0, "max": 153.0 }, "score": { "n": 8, "mean": 2.0, "min": 0.0, "max": 3.0 }, "survival_steps": { "n": 8, "mean": 15.625, "min": 6.0, "max": 35.0 }, "target_food": { "n": 8, "mean": 3.0, "min": 3.0, "max": 3.0 } } }, "ood/basic": { "n": 128, "successes": 126, "success_rate": 0.984375, "wilson95": [ 0.9448190233201827, 0.9957045502898559 ], "mean_steps": 5.3203125, "mean_decisions": 5.3203125, "metrics": { "ammo": { "n": 128, "mean": 48.046875, "min": 30.0, "max": 49.0 }, "ammo_consumed": { "n": 128, "mean": 1.953125, "min": 1.0, "max": 20.0 }, "damage_dealt": { "n": 128, "mean": 9.296875, "min": 0.0, "max": 15.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 5.3203125, "min": 1.0, "max": 36.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.984375, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.984375, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": 55.1953125, "min": -386.0, "max": 96.0 }, "physical_ticks": { "n": 128, "mean": 39.3828125, "min": 5.0, "max": 286.0 } } }, "ood/predict_position": { "n": 128, "successes": 9, "success_rate": 0.0703125, "wilson95": [ 0.03742976830393997, 0.1282348034936063 ], "mean_steps": 18.5625, "mean_decisions": 18.5625, "metrics": { "ammo": { "n": 128, "mean": 0.0546875, "min": 0.0, "max": 1.0 }, "ammo_consumed": { "n": 128, "mean": 0.9453125, "min": 0.0, "max": 1.0 }, "damage_dealt": { "n": 128, "mean": 14.8046875, "min": 0.0, "max": 196.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 18.5625, "min": 7.0, "max": 36.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.0703125, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.0703125, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": -0.07669531250000007, "min": -0.2840000000000002, "max": 0.948 }, "physical_ticks": { "n": 128, "mean": 147.0078125, "min": 52.0, "max": 284.0 } } } }, "macro_by_split": { "test": { "success_rate": 0.5643229166666667, "weights": { "maze": 0.3333333333333333, "snake": 0.3333333333333333, "basic": 0.16666666666666666, "predict_position": 0.16666666666666666 }, "interval": null }, "ood": { "success_rate": 0.40911458333333334, "weights": { "maze": 0.3333333333333333, "snake": 0.3333333333333333, "basic": 0.16666666666666666, "predict_position": 0.16666666666666666 }, "interval": null } }, "paired_vs_jev": { "test/maze": { "n": 10, "wins_vs_jev": 0, "losses_vs_jev": 3, "both_success": 4, "both_failure": 3, "success_rate_delta_vs_jev": -0.3, "exact_mcnemar_two_sided_p": 0.25, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [ "test-maze16-7830000", "test-maze8-308040", "test-maze8-308053" ] }, "test/snake": { "n": 8, "wins_vs_jev": 0, "losses_vs_jev": 2, "both_success": 6, "both_failure": 0, "success_rate_delta_vs_jev": -0.25, "exact_mcnemar_two_sided_p": 0.5, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [ "test-snake12-7832000", "test-snake12-7832001" ] }, "test/basic": { "n": 128, "wins_vs_jev": 72, "losses_vs_jev": 0, "both_success": 56, "both_failure": 0, "success_rate_delta_vs_jev": 0.5625, "exact_mcnemar_two_sided_p": 4.235164736271502e-22, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [ "test-appo_basic-9030000", "test-appo_basic-9030001", "test-appo_basic-9030002", "test-appo_basic-9030003", "test-appo_basic-9030005", "test-appo_basic-9030009", "test-appo_basic-9030010", "test-appo_basic-9030011", "test-appo_basic-9030012", "test-appo_basic-9030015", "test-appo_basic-9030017", "test-appo_basic-9030018", "test-appo_basic-9030019", "test-appo_basic-9030020", "test-appo_basic-9030021", "test-appo_basic-9030023", "test-appo_basic-9030027", "test-appo_basic-9030028", "test-appo_basic-9030029", "test-appo_basic-9030030", "test-appo_basic-9030032", "test-appo_basic-9030033", "test-appo_basic-9030035", "test-appo_basic-9030041", "test-appo_basic-9030043", "test-appo_basic-9030044", "test-appo_basic-9030045", "test-appo_basic-9030046", "test-appo_basic-9030047", "test-appo_basic-9030048", "test-appo_basic-9030051", "test-appo_basic-9030052", "test-appo_basic-9030053", "test-appo_basic-9030054", "test-appo_basic-9030055", "test-appo_basic-9030056", "test-appo_basic-9030057", "test-appo_basic-9030059", "test-appo_basic-9030060", "test-appo_basic-9030064", "test-appo_basic-9030066", "test-appo_basic-9030069", "test-appo_basic-9030070", "test-appo_basic-9030071", "test-appo_basic-9030073", "test-appo_basic-9030074", "test-appo_basic-9030075", "test-appo_basic-9030076", "test-appo_basic-9030079", "test-appo_basic-9030081", "test-appo_basic-9030082", "test-appo_basic-9030084", "test-appo_basic-9030085", "test-appo_basic-9030087", "test-appo_basic-9030089", "test-appo_basic-9030091", "test-appo_basic-9030094", "test-appo_basic-9030097", "test-appo_basic-9030098", "test-appo_basic-9030099", "test-appo_basic-9030102", "test-appo_basic-9030103", "test-appo_basic-9030105", "test-appo_basic-9030110", "test-appo_basic-9030115", "test-appo_basic-9030116", "test-appo_basic-9030117", "test-appo_basic-9030118", "test-appo_basic-9030121", "test-appo_basic-9030124", "test-appo_basic-9030125", "test-appo_basic-9030126" ], "loss_case_ids": [] }, "test/predict_position": { "n": 128, "wins_vs_jev": 0, "losses_vs_jev": 0, "both_success": 11, "both_failure": 117, "success_rate_delta_vs_jev": 0.0, "exact_mcnemar_two_sided_p": 1.0, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [] }, "ood/maze": { "n": 10, "wins_vs_jev": 1, "losses_vs_jev": 2, "both_success": 1, "both_failure": 6, "success_rate_delta_vs_jev": -0.1, "exact_mcnemar_two_sided_p": 1.0, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [ "ood-maze8-408000" ], "loss_case_ids": [ "ood-maze16-7840001", "ood-maze8-408004" ] }, "ood/snake": { "n": 8, "wins_vs_jev": 0, "losses_vs_jev": 2, "both_success": 4, "both_failure": 2, "success_rate_delta_vs_jev": -0.25, "exact_mcnemar_two_sided_p": 0.5, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [ "ood-snake12-7842000", "ood-snake8-418000" ] }, "ood/basic": { "n": 128, "wins_vs_jev": 67, "losses_vs_jev": 0, "both_success": 59, "both_failure": 2, "success_rate_delta_vs_jev": 0.5234375, "exact_mcnemar_two_sided_p": 1.3552527156068805e-20, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [ "ood-appo_basic-9040000", "ood-appo_basic-9040001", "ood-appo_basic-9040002", "ood-appo_basic-9040003", "ood-appo_basic-9040004", "ood-appo_basic-9040006", "ood-appo_basic-9040008", "ood-appo_basic-9040011", "ood-appo_basic-9040013", "ood-appo_basic-9040014", "ood-appo_basic-9040019", "ood-appo_basic-9040022", "ood-appo_basic-9040023", "ood-appo_basic-9040024", "ood-appo_basic-9040026", "ood-appo_basic-9040029", "ood-appo_basic-9040030", "ood-appo_basic-9040031", "ood-appo_basic-9040032", "ood-appo_basic-9040033", "ood-appo_basic-9040036", "ood-appo_basic-9040038", "ood-appo_basic-9040039", "ood-appo_basic-9040040", "ood-appo_basic-9040041", "ood-appo_basic-9040042", "ood-appo_basic-9040044", "ood-appo_basic-9040045", "ood-appo_basic-9040050", "ood-appo_basic-9040051", "ood-appo_basic-9040052", "ood-appo_basic-9040053", "ood-appo_basic-9040056", "ood-appo_basic-9040057", "ood-appo_basic-9040059", "ood-appo_basic-9040060", "ood-appo_basic-9040061", "ood-appo_basic-9040064", "ood-appo_basic-9040066", "ood-appo_basic-9040067", "ood-appo_basic-9040068", "ood-appo_basic-9040071", "ood-appo_basic-9040076", "ood-appo_basic-9040078", "ood-appo_basic-9040079", "ood-appo_basic-9040083", "ood-appo_basic-9040086", "ood-appo_basic-9040090", "ood-appo_basic-9040091", "ood-appo_basic-9040092", "ood-appo_basic-9040093", "ood-appo_basic-9040094", "ood-appo_basic-9040095", "ood-appo_basic-9040097", "ood-appo_basic-9040100", "ood-appo_basic-9040102", "ood-appo_basic-9040103", "ood-appo_basic-9040104", "ood-appo_basic-9040111", "ood-appo_basic-9040114", "ood-appo_basic-9040115", "ood-appo_basic-9040116", "ood-appo_basic-9040117", "ood-appo_basic-9040118", "ood-appo_basic-9040123", "ood-appo_basic-9040124", "ood-appo_basic-9040126" ], "loss_case_ids": [] }, "ood/predict_position": { "n": 128, "wins_vs_jev": 1, "losses_vs_jev": 0, "both_success": 8, "both_failure": 119, "success_rate_delta_vs_jev": 0.0078125, "exact_mcnemar_two_sided_p": 1.0, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [ "ood-sonic_predict_position-9300827" ], "loss_case_ids": [] } }, "macro_delta_vs_jev": { "test": -0.08958333333333332, "ood": -0.028125 } }, "selected": { "source": { "episodes_path": "runs/sonic_unified_sft_v1/experiment/selected_test.jsonl", "manifest_path": "runs/sonic_unified_sft_v1/experiment/selected_test.manifest.json", "episodes_sha256": "0a6fd7e0ce9498472add6277a869d4b9e04bcb366c3241ed823993f3d21b6077", "manifest_sha256": "ca3cf4102b0a1f67d3d073020ff7eb010234021938a4e7d9952e8d8117a617e3", "cases_sha256": "d00f5b6aeaa9f2aac3e4b5010b31df4e86f77674cb588e17bc3f7d853e4a1f36", "continuation_policy_id": "ab8a8c8452b17c18c47001d47b1fd6bd7c7fed11842c8ffabe11fdde200a9962", "policy": { "engine": "checkpoint", "controller": "greedy", "epsilon": 0.1, "sampling_seed": 17, "temperature": 1.0, "source_sha256": { "unified_game_pipeline.py": "c176e06d7088bafbf0fbe4ca7e2be05eef6f1f5c3bf00b2d946caa21cce42313", "unified_grid_envs.py": "8ab8a9ff56f88866efc26bca41c16416e432d5ed8415cdcd33f622c5c7030663", "unified_doom_env.py": "fba56e3e97fe7bf0940962814308f63fcd2a788bf930e36f7f75b2aaa1889039", "scaled_maze.py": "4eef4a1052c913755ccccc54b9e7ee2cca5e6a1258ab748d59bfd45386d410bb", "snake_game.py": "adacc7a6013f0195e5a0952bc2a13ec14964325b52b56b2e69d780794edf549e", "evaluate_composed_maze.py": "b443469cedf81b95c154e33a7d90accd9bb7cd7423a40bfd7314be1f42fc9ae5", "predict_toy_decisions.py": "6eedc49aaf9a81763677e3cff4e87fc4e499bfc5decf0bb4678f452f578dd92b", "train_toy_decisions.py": "4f39babd5575e43d7acf3eca357329a93c60041f334bc2bd7424929bf9c0da86", "unified_jev_worker.mjs": "e844d42fcf6f1ac844c7525746cef3b8d87848682ccbaafe91e1c12294bc1dbf", "teachers.mjs": "6f8dd533e34a8f491530a26197081461e70977fce1857924ac5856c4817787ab" }, "tie_break": "lexicographic_first", "environment_contract": "finite_task_deadline_v1", "checkpoint_sha256": { "config.json": "901038b5f1785c748c3fb4d35131e910951f2178f9a8bf024ff57aca556a30e9", "best.safetensors": "f68c47d66998231b86b7e91b4ed5e82ae23acf104c8b7cd6d165c3ac7b7ffe1b", "tokenizer/tokenizer.json": "be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506", "backbone_config/config.json": "30c011854471509747858ac07ffb8a4dab2bc6c04035b184729b69500c557c54" } } }, "validation": { "matched_episodes": 548, "controller_transitions_checked": 11948, "physical_replay_performed_by_this_script": false }, "by_split_category": { "test/maze": { "n": 10, "successes": 4, "success_rate": 0.4, "wilson95": [ 0.16818032970623617, 0.6873262302663417 ], "mean_steps": 184.9, "mean_decisions": 150.4, "metrics": { "collisions": { "n": 10, "mean": 90.4, "min": 7.0, "max": 355.0 }, "decision_steps": { "n": 10, "mean": 184.9, "min": 21.0, "max": 700.0 }, "physical_steps": { "n": 10, "mean": 208.5, "min": 21.0, "max": 800.0 }, "remaining_steps": { "n": 10, "mean": 2.7, "min": 0.0, "max": 11.0 }, "reposition_steps": { "n": 10, "mean": 23.6, "min": 0.0, "max": 110.0 }, "visited_cells": { "n": 10, "mean": 95.5, "min": 14.0, "max": 371.0 } } }, "test/snake": { "n": 8, "successes": 8, "success_rate": 1.0, "wilson95": [ 0.6755924351161198, 1.0 ], "mean_steps": 20.25, "mean_decisions": 20.25, "metrics": { "body_length": { "n": 8, "mean": 5.25, "min": 5.0, "max": 6.0 }, "collisions": { "n": 8, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decision_steps": { "n": 8, "mean": 20.25, "min": 4.0, "max": 44.0 }, "food_collected": { "n": 8, "mean": 2.25, "min": 2.0, "max": 3.0 }, "initial_score": { "n": 8, "mean": 0.0, "min": 0.0, "max": 0.0 }, "physical_steps": { "n": 8, "mean": 20.25, "min": 4.0, "max": 44.0 }, "remaining_steps": { "n": 8, "mean": 91.75, "min": 64.0, "max": 120.0 }, "score": { "n": 8, "mean": 2.25, "min": 2.0, "max": 3.0 }, "survival_steps": { "n": 8, "mean": 20.25, "min": 4.0, "max": 44.0 }, "target_food": { "n": 8, "mean": 2.25, "min": 2.0, "max": 3.0 } } }, "test/basic": { "n": 128, "successes": 128, "success_rate": 1.0, "wilson95": [ 0.9708630437264916, 1.0 ], "mean_steps": 6.140625, "mean_decisions": 6.140625, "metrics": { "ammo": { "n": 128, "mean": 48.8359375, "min": 47.0, "max": 49.0 }, "ammo_consumed": { "n": 128, "mean": 1.1640625, "min": 1.0, "max": 3.0 }, "damage_dealt": { "n": 128, "mean": 9.765625, "min": 5.0, "max": 15.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 6.140625, "min": 2.0, "max": 14.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 1.0, "min": 1.0, "max": 1.0 }, "kills": { "n": 128, "mean": 1.0, "min": 1.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": 78.453125, "min": 38.0, "max": 96.0 }, "physical_ticks": { "n": 128, "mean": 21.7265625, "min": 5.0, "max": 53.0 } } }, "test/predict_position": { "n": 128, "successes": 27, "success_rate": 0.2109375, "wilson95": [ 0.14921308750498036, 0.2895067153406417 ], "mean_steps": 25.296875, "mean_decisions": 25.296875, "metrics": { "ammo": { "n": 128, "mean": 0.0078125, "min": 0.0, "max": 1.0 }, "ammo_consumed": { "n": 128, "mean": 0.9921875, "min": 0.0, "max": 1.0 }, "damage_dealt": { "n": 128, "mean": 39.7265625, "min": 0.0, "max": 216.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 25.296875, "min": 11.0, "max": 71.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.2109375, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.2109375, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": 0.11085156249999993, "min": -0.2840000000000002, "max": 0.958 }, "physical_ticks": { "n": 128, "mean": 100.0859375, "min": 42.0, "max": 284.0 } } }, "ood/maze": { "n": 10, "successes": 2, "success_rate": 0.2, "wilson95": [ 0.05668215145437522, 0.5098375284633582 ], "mean_steps": 260.5, "mean_decisions": 216.5, "metrics": { "collisions": { "n": 10, "mean": 122.8, "min": 65.0, "max": 362.0 }, "decision_steps": { "n": 10, "mean": 260.5, "min": 142.0, "max": 724.0 }, "physical_steps": { "n": 10, "mean": 285.1, "min": 144.0, "max": 800.0 }, "remaining_steps": { "n": 10, "mean": 2.9, "min": 0.0, "max": 16.0 }, "reposition_steps": { "n": 10, "mean": 24.6, "min": 0.0, "max": 106.0 }, "visited_cells": { "n": 10, "mean": 138.7, "min": 68.0, "max": 450.0 } } }, "ood/snake": { "n": 8, "successes": 5, "success_rate": 0.625, "wilson95": [ 0.3057423946026273, 0.8631557141764027 ], "mean_steps": 16.5, "mean_decisions": 16.5, "metrics": { "body_length": { "n": 8, "mean": 5.125, "min": 3.0, "max": 6.0 }, "collisions": { "n": 8, "mean": 0.375, "min": 0.0, "max": 1.0 }, "decision_steps": { "n": 8, "mean": 16.5, "min": 7.0, "max": 35.0 }, "food_collected": { "n": 8, "mean": 2.125, "min": 0.0, "max": 3.0 }, "initial_score": { "n": 8, "mean": 0.0, "min": 0.0, "max": 0.0 }, "physical_steps": { "n": 8, "mean": 16.5, "min": 7.0, "max": 35.0 }, "remaining_steps": { "n": 8, "mean": 143.5, "min": 125.0, "max": 153.0 }, "score": { "n": 8, "mean": 2.125, "min": 0.0, "max": 3.0 }, "survival_steps": { "n": 8, "mean": 16.125, "min": 6.0, "max": 35.0 }, "target_food": { "n": 8, "mean": 3.0, "min": 3.0, "max": 3.0 } } }, "ood/basic": { "n": 128, "successes": 128, "success_rate": 1.0, "wilson95": [ 0.9708630437264916, 1.0 ], "mean_steps": 4.7265625, "mean_decisions": 4.7265625, "metrics": { "ammo": { "n": 128, "mean": 48.4765625, "min": 43.0, "max": 49.0 }, "ammo_consumed": { "n": 128, "mean": 1.5234375, "min": 1.0, "max": 7.0 }, "damage_dealt": { "n": 128, "mean": 9.453125, "min": 5.0, "max": 15.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 4.7265625, "min": 1.0, "max": 21.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 1.0, "min": 1.0, "max": 1.0 }, "kills": { "n": 128, "mean": 1.0, "min": 1.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": 63.7109375, "min": -92.0, "max": 96.0 }, "physical_ticks": { "n": 128, "mean": 34.671875, "min": 5.0, "max": 163.0 } } }, "ood/predict_position": { "n": 128, "successes": 10, "success_rate": 0.078125, "wilson95": [ 0.04298737505098223, 0.1378469318047905 ], "mean_steps": 20.0859375, "mean_decisions": 20.0859375, "metrics": { "ammo": { "n": 128, "mean": 0.1015625, "min": 0.0, "max": 1.0 }, "ammo_consumed": { "n": 128, "mean": 0.8984375, "min": 0.0, "max": 1.0 }, "damage_dealt": { "n": 128, "mean": 17.25, "min": 0.0, "max": 214.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 20.0859375, "min": 6.0, "max": 36.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.078125, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.078125, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": -0.08084375000000009, "min": -0.2840000000000002, "max": 0.955 }, "physical_ticks": { "n": 128, "mean": 158.96875, "min": 45.0, "max": 284.0 } } } }, "macro_by_split": { "test": { "success_rate": 0.6684895833333333, "weights": { "maze": 0.3333333333333333, "snake": 0.3333333333333333, "basic": 0.16666666666666666, "predict_position": 0.16666666666666666 }, "interval": null }, "ood": { "success_rate": 0.45468749999999997, "weights": { "maze": 0.3333333333333333, "snake": 0.3333333333333333, "basic": 0.16666666666666666, "predict_position": 0.16666666666666666 }, "interval": null } }, "paired_vs_jev": { "test/maze": { "n": 10, "wins_vs_jev": 0, "losses_vs_jev": 3, "both_success": 4, "both_failure": 3, "success_rate_delta_vs_jev": -0.3, "exact_mcnemar_two_sided_p": 0.25, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [ "test-maze16-7830000", "test-maze8-308040", "test-maze8-308053" ] }, "test/snake": { "n": 8, "wins_vs_jev": 0, "losses_vs_jev": 0, "both_success": 8, "both_failure": 0, "success_rate_delta_vs_jev": 0.0, "exact_mcnemar_two_sided_p": 1.0, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [] }, "test/basic": { "n": 128, "wins_vs_jev": 72, "losses_vs_jev": 0, "both_success": 56, "both_failure": 0, "success_rate_delta_vs_jev": 0.5625, "exact_mcnemar_two_sided_p": 4.235164736271502e-22, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [ "test-appo_basic-9030000", "test-appo_basic-9030001", "test-appo_basic-9030002", "test-appo_basic-9030003", "test-appo_basic-9030005", "test-appo_basic-9030009", "test-appo_basic-9030010", "test-appo_basic-9030011", "test-appo_basic-9030012", "test-appo_basic-9030015", "test-appo_basic-9030017", "test-appo_basic-9030018", "test-appo_basic-9030019", "test-appo_basic-9030020", "test-appo_basic-9030021", "test-appo_basic-9030023", "test-appo_basic-9030027", "test-appo_basic-9030028", "test-appo_basic-9030029", "test-appo_basic-9030030", "test-appo_basic-9030032", "test-appo_basic-9030033", "test-appo_basic-9030035", "test-appo_basic-9030041", "test-appo_basic-9030043", "test-appo_basic-9030044", "test-appo_basic-9030045", "test-appo_basic-9030046", "test-appo_basic-9030047", "test-appo_basic-9030048", "test-appo_basic-9030051", "test-appo_basic-9030052", "test-appo_basic-9030053", "test-appo_basic-9030054", "test-appo_basic-9030055", "test-appo_basic-9030056", "test-appo_basic-9030057", "test-appo_basic-9030059", "test-appo_basic-9030060", "test-appo_basic-9030064", "test-appo_basic-9030066", "test-appo_basic-9030069", "test-appo_basic-9030070", "test-appo_basic-9030071", "test-appo_basic-9030073", "test-appo_basic-9030074", "test-appo_basic-9030075", "test-appo_basic-9030076", "test-appo_basic-9030079", "test-appo_basic-9030081", "test-appo_basic-9030082", "test-appo_basic-9030084", "test-appo_basic-9030085", "test-appo_basic-9030087", "test-appo_basic-9030089", "test-appo_basic-9030091", "test-appo_basic-9030094", "test-appo_basic-9030097", "test-appo_basic-9030098", "test-appo_basic-9030099", "test-appo_basic-9030102", "test-appo_basic-9030103", "test-appo_basic-9030105", "test-appo_basic-9030110", "test-appo_basic-9030115", "test-appo_basic-9030116", "test-appo_basic-9030117", "test-appo_basic-9030118", "test-appo_basic-9030121", "test-appo_basic-9030124", "test-appo_basic-9030125", "test-appo_basic-9030126" ], "loss_case_ids": [] }, "test/predict_position": { "n": 128, "wins_vs_jev": 25, "losses_vs_jev": 9, "both_success": 2, "both_failure": 92, "success_rate_delta_vs_jev": 0.125, "exact_mcnemar_two_sided_p": 0.00904118549078703, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [ "test-sonic_predict_position-9300641", "test-sonic_predict_position-9300645", "test-sonic_predict_position-9300647", "test-sonic_predict_position-9300657", "test-sonic_predict_position-9300666", "test-sonic_predict_position-9300667", "test-sonic_predict_position-9300673", "test-sonic_predict_position-9300677", "test-sonic_predict_position-9300679", "test-sonic_predict_position-9300683", "test-sonic_predict_position-9300690", "test-sonic_predict_position-9300692", "test-sonic_predict_position-9300693", "test-sonic_predict_position-9300699", "test-sonic_predict_position-9300717", "test-sonic_predict_position-9300720", "test-sonic_predict_position-9300722", "test-sonic_predict_position-9300725", "test-sonic_predict_position-9300728", "test-sonic_predict_position-9300730", "test-sonic_predict_position-9300738", "test-sonic_predict_position-9300747", "test-sonic_predict_position-9300753", "test-sonic_predict_position-9300755", "test-sonic_predict_position-9300766" ], "loss_case_ids": [ "test-sonic_predict_position-9300644", "test-sonic_predict_position-9300655", "test-sonic_predict_position-9300672", "test-sonic_predict_position-9300678", "test-sonic_predict_position-9300687", "test-sonic_predict_position-9300714", "test-sonic_predict_position-9300721", "test-sonic_predict_position-9300727", "test-sonic_predict_position-9300748" ] }, "ood/maze": { "n": 10, "wins_vs_jev": 0, "losses_vs_jev": 1, "both_success": 2, "both_failure": 7, "success_rate_delta_vs_jev": -0.1, "exact_mcnemar_two_sided_p": 1.0, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [ "ood-maze16-7840001" ] }, "ood/snake": { "n": 8, "wins_vs_jev": 0, "losses_vs_jev": 1, "both_success": 5, "both_failure": 2, "success_rate_delta_vs_jev": -0.125, "exact_mcnemar_two_sided_p": 1.0, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [ "ood-snake12-7842000" ] }, "ood/basic": { "n": 128, "wins_vs_jev": 69, "losses_vs_jev": 0, "both_success": 59, "both_failure": 0, "success_rate_delta_vs_jev": 0.5390625, "exact_mcnemar_two_sided_p": 3.3881317890172014e-21, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [ "ood-appo_basic-9040000", "ood-appo_basic-9040001", "ood-appo_basic-9040002", "ood-appo_basic-9040003", "ood-appo_basic-9040004", "ood-appo_basic-9040005", "ood-appo_basic-9040006", "ood-appo_basic-9040008", "ood-appo_basic-9040011", "ood-appo_basic-9040013", "ood-appo_basic-9040014", "ood-appo_basic-9040019", "ood-appo_basic-9040022", "ood-appo_basic-9040023", "ood-appo_basic-9040024", "ood-appo_basic-9040026", "ood-appo_basic-9040029", "ood-appo_basic-9040030", "ood-appo_basic-9040031", "ood-appo_basic-9040032", "ood-appo_basic-9040033", "ood-appo_basic-9040036", "ood-appo_basic-9040038", "ood-appo_basic-9040039", "ood-appo_basic-9040040", "ood-appo_basic-9040041", "ood-appo_basic-9040042", "ood-appo_basic-9040044", "ood-appo_basic-9040045", "ood-appo_basic-9040050", "ood-appo_basic-9040051", "ood-appo_basic-9040052", "ood-appo_basic-9040053", "ood-appo_basic-9040056", "ood-appo_basic-9040057", "ood-appo_basic-9040059", "ood-appo_basic-9040060", "ood-appo_basic-9040061", "ood-appo_basic-9040064", "ood-appo_basic-9040066", "ood-appo_basic-9040067", "ood-appo_basic-9040068", "ood-appo_basic-9040071", "ood-appo_basic-9040076", "ood-appo_basic-9040078", "ood-appo_basic-9040079", "ood-appo_basic-9040083", "ood-appo_basic-9040086", "ood-appo_basic-9040087", "ood-appo_basic-9040090", "ood-appo_basic-9040091", "ood-appo_basic-9040092", "ood-appo_basic-9040093", "ood-appo_basic-9040094", "ood-appo_basic-9040095", "ood-appo_basic-9040097", "ood-appo_basic-9040100", "ood-appo_basic-9040102", "ood-appo_basic-9040103", "ood-appo_basic-9040104", "ood-appo_basic-9040111", "ood-appo_basic-9040114", "ood-appo_basic-9040115", "ood-appo_basic-9040116", "ood-appo_basic-9040117", "ood-appo_basic-9040118", "ood-appo_basic-9040123", "ood-appo_basic-9040124", "ood-appo_basic-9040126" ], "loss_case_ids": [] }, "ood/predict_position": { "n": 128, "wins_vs_jev": 10, "losses_vs_jev": 8, "both_success": 0, "both_failure": 110, "success_rate_delta_vs_jev": 0.015625, "exact_mcnemar_two_sided_p": 0.8145294189453125, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [ "ood-sonic_predict_position-9300773", "ood-sonic_predict_position-9300775", "ood-sonic_predict_position-9300776", "ood-sonic_predict_position-9300782", "ood-sonic_predict_position-9300812", "ood-sonic_predict_position-9300815", "ood-sonic_predict_position-9300817", "ood-sonic_predict_position-9300843", "ood-sonic_predict_position-9300870", "ood-sonic_predict_position-9300881" ], "loss_case_ids": [ "ood-sonic_predict_position-9300787", "ood-sonic_predict_position-9300790", "ood-sonic_predict_position-9300793", "ood-sonic_predict_position-9300811", "ood-sonic_predict_position-9300825", "ood-sonic_predict_position-9300844", "ood-sonic_predict_position-9300861", "ood-sonic_predict_position-9300884" ] } }, "macro_delta_vs_jev": { "test": 0.01458333333333334, "ood": 0.01744791666666667 } }, "jev": { "source": { "episodes_path": "runs/sonic_unified_sft_v1/jev_parallel_recovery_v2/episodes.jsonl", "manifest_path": "runs/sonic_unified_sft_v1/jev_parallel_recovery_v2/episodes.manifest.json", "episodes_sha256": "f5d54efe726e737cd394b32b11bde81b59e72fa107e2ba00e9b44413728fb945", "manifest_sha256": "7828ab10434b1cedf6ba4c46b6f4b0a9680512ddfac00362ee036d85d827d248", "cases_sha256": "d00f5b6aeaa9f2aac3e4b5010b31df4e86f77674cb588e17bc3f7d853e4a1f36", "continuation_policy_id": "f3644f9c0e364b6bb3a499f6a5c0b9810aed573992476ca94f0f345a3fd710a6", "policy": { "engine": "jev", "controller": "greedy", "epsilon": 0.1, "sampling_seed": 17, "temperature": 1.0, "source_sha256": { "unified_game_pipeline.py": "c176e06d7088bafbf0fbe4ca7e2be05eef6f1f5c3bf00b2d946caa21cce42313", "unified_grid_envs.py": "8ab8a9ff56f88866efc26bca41c16416e432d5ed8415cdcd33f622c5c7030663", "unified_doom_env.py": "fba56e3e97fe7bf0940962814308f63fcd2a788bf930e36f7f75b2aaa1889039", "scaled_maze.py": "4eef4a1052c913755ccccc54b9e7ee2cca5e6a1258ab748d59bfd45386d410bb", "snake_game.py": "adacc7a6013f0195e5a0952bc2a13ec14964325b52b56b2e69d780794edf549e", "evaluate_composed_maze.py": "b443469cedf81b95c154e33a7d90accd9bb7cd7423a40bfd7314be1f42fc9ae5", "predict_toy_decisions.py": "6eedc49aaf9a81763677e3cff4e87fc4e499bfc5decf0bb4678f452f578dd92b", "train_toy_decisions.py": "4f39babd5575e43d7acf3eca357329a93c60041f334bc2bd7424929bf9c0da86", "unified_jev_worker.mjs": "e844d42fcf6f1ac844c7525746cef3b8d87848682ccbaafe91e1c12294bc1dbf", "teachers.mjs": "6f8dd533e34a8f491530a26197081461e70977fce1857924ac5856c4817787ab" }, "tie_break": "lexicographic_first", "environment_contract": "finite_task_deadline_v1", "model": "typesafe-ai/jev" } }, "validation": { "matched_episodes": 548, "controller_transitions_checked": 19934, "physical_replay_performed_by_this_script": false }, "by_split_category": { "test/maze": { "n": 10, "successes": 7, "success_rate": 0.7, "wilson95": [ 0.3967781474611453, 0.8922087325936988 ], "mean_steps": 188.6, "mean_decisions": 155.7, "metrics": { "collisions": { "n": 10, "mean": 89.1, "min": 10.0, "max": 345.0 }, "decision_steps": { "n": 10, "mean": 188.6, "min": 24.0, "max": 727.0 }, "physical_steps": { "n": 10, "mean": 208.3, "min": 24.0, "max": 800.0 }, "remaining_steps": { "n": 10, "mean": 2.9, "min": 0.0, "max": 8.0 }, "reposition_steps": { "n": 10, "mean": 19.7, "min": 0.0, "max": 108.0 }, "visited_cells": { "n": 10, "mean": 100.5, "min": 15.0, "max": 383.0 } } }, "test/snake": { "n": 8, "successes": 8, "success_rate": 1.0, "wilson95": [ 0.6755924351161198, 1.0 ], "mean_steps": 41.875, "mean_decisions": 41.875, "metrics": { "body_length": { "n": 8, "mean": 5.25, "min": 5.0, "max": 6.0 }, "collisions": { "n": 8, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decision_steps": { "n": 8, "mean": 41.875, "min": 4.0, "max": 136.0 }, "food_collected": { "n": 8, "mean": 2.25, "min": 2.0, "max": 3.0 }, "initial_score": { "n": 8, "mean": 0.0, "min": 0.0, "max": 0.0 }, "physical_steps": { "n": 8, "mean": 41.875, "min": 4.0, "max": 136.0 }, "remaining_steps": { "n": 8, "mean": 70.125, "min": 24.0, "max": 92.0 }, "score": { "n": 8, "mean": 2.25, "min": 2.0, "max": 3.0 }, "survival_steps": { "n": 8, "mean": 41.875, "min": 4.0, "max": 136.0 }, "target_food": { "n": 8, "mean": 2.25, "min": 2.0, "max": 3.0 } } }, "test/basic": { "n": 128, "successes": 56, "success_rate": 0.4375, "wilson95": [ 0.3546231312705304, 0.5240189882636581 ], "mean_steps": 47.5, "mean_decisions": 47.5, "metrics": { "ammo": { "n": 128, "mean": 36.5546875, "min": 29.0, "max": 49.0 }, "ammo_consumed": { "n": 128, "mean": 13.4453125, "min": 1.0, "max": 21.0 }, "damage_dealt": { "n": 128, "mean": 4.4140625, "min": 0.0, "max": 15.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 47.5, "min": 2.0, "max": 72.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.4375, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.4375, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": -208.7421875, "min": -391.0, "max": 96.0 }, "physical_ticks": { "n": 128, "mean": 187.890625, "min": 5.0, "max": 286.0 } } }, "test/predict_position": { "n": 128, "successes": 11, "success_rate": 0.0859375, "wilson95": [ 0.0486631313713862, 0.14734091054261297 ], "mean_steps": 26.640625, "mean_decisions": 26.640625, "metrics": { "ammo": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "ammo_consumed": { "n": 128, "mean": 1.0, "min": 1.0, "max": 1.0 }, "damage_dealt": { "n": 128, "mean": 16.2421875, "min": 0.0, "max": 208.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 26.640625, "min": 10.0, "max": 71.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.0859375, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.0859375, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": -0.019578125000000068, "min": -0.2840000000000002, "max": 0.962 }, "physical_ticks": { "n": 128, "mean": 105.515625, "min": 38.0, "max": 284.0 } } }, "ood/maze": { "n": 10, "successes": 3, "success_rate": 0.3, "wilson95": [ 0.10779126740630099, 0.6032218525388546 ], "mean_steps": 255.5, "mean_decisions": 207.4, "metrics": { "collisions": { "n": 10, "mean": 126.8, "min": 51.0, "max": 357.0 }, "decision_steps": { "n": 10, "mean": 255.5, "min": 129.0, "max": 691.0 }, "physical_steps": { "n": 10, "mean": 282.6, "min": 129.0, "max": 800.0 }, "remaining_steps": { "n": 10, "mean": 5.4, "min": 0.0, "max": 31.0 }, "reposition_steps": { "n": 10, "mean": 27.1, "min": 0.0, "max": 112.0 }, "visited_cells": { "n": 10, "mean": 129.7, "min": 69.0, "max": 348.0 } } }, "ood/snake": { "n": 8, "successes": 6, "success_rate": 0.75, "wilson95": [ 0.40927543031016894, 0.928520787247891 ], "mean_steps": 23.25, "mean_decisions": 23.25, "metrics": { "body_length": { "n": 8, "mean": 5.5, "min": 4.0, "max": 6.0 }, "collisions": { "n": 8, "mean": 0.25, "min": 0.0, "max": 1.0 }, "decision_steps": { "n": 8, "mean": 23.25, "min": 7.0, "max": 59.0 }, "food_collected": { "n": 8, "mean": 2.5, "min": 1.0, "max": 3.0 }, "initial_score": { "n": 8, "mean": 0.0, "min": 0.0, "max": 0.0 }, "physical_steps": { "n": 8, "mean": 23.25, "min": 7.0, "max": 59.0 }, "remaining_steps": { "n": 8, "mean": 136.75, "min": 101.0, "max": 153.0 }, "score": { "n": 8, "mean": 2.5, "min": 1.0, "max": 3.0 }, "survival_steps": { "n": 8, "mean": 23.0, "min": 6.0, "max": 59.0 }, "target_food": { "n": 8, "mean": 3.0, "min": 3.0, "max": 3.0 } } }, "ood/basic": { "n": 128, "successes": 59, "success_rate": 0.4609375, "wilson95": [ 0.3769811559809195, 0.5471701687279482 ], "mean_steps": 24.4296875, "mean_decisions": 24.4296875, "metrics": { "ammo": { "n": 128, "mean": 36.3046875, "min": 29.0, "max": 49.0 }, "ammo_consumed": { "n": 128, "mean": 13.6953125, "min": 1.0, "max": 21.0 }, "damage_dealt": { "n": 128, "mean": 4.84375, "min": 0.0, "max": 15.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 24.4296875, "min": 1.0, "max": 36.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.4609375, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.4609375, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": -212.296875, "min": -391.0, "max": 96.0 }, "physical_ticks": { "n": 128, "mean": 192.6796875, "min": 5.0, "max": 286.0 } } }, "ood/predict_position": { "n": 128, "successes": 8, "success_rate": 0.0625, "wilson95": [ 0.032006925695061195, 0.11848791104425867 ], "mean_steps": 18.3984375, "mean_decisions": 18.3984375, "metrics": { "ammo": { "n": 128, "mean": 0.0546875, "min": 0.0, "max": 1.0 }, "ammo_consumed": { "n": 128, "mean": 0.9453125, "min": 0.0, "max": 1.0 }, "damage_dealt": { "n": 128, "mean": 13.4296875, "min": 0.0, "max": 196.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 18.3984375, "min": 7.0, "max": 36.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.0625, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.0625, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": -0.08321875000000008, "min": -0.2840000000000002, "max": 0.948 }, "physical_ticks": { "n": 128, "mean": 145.71875, "min": 52.0, "max": 284.0 } } } }, "macro_by_split": { "test": { "success_rate": 0.6539062499999999, "weights": { "maze": 0.3333333333333333, "snake": 0.3333333333333333, "basic": 0.16666666666666666, "predict_position": 0.16666666666666666 }, "interval": null }, "ood": { "success_rate": 0.4372395833333333, "weights": { "maze": 0.3333333333333333, "snake": 0.3333333333333333, "basic": 0.16666666666666666, "predict_position": 0.16666666666666666 }, "interval": null } } }, "untuned_qwen": { "source": { "episodes_path": "runs/sonic_unified_sft_v1/native_test.jsonl", "manifest_path": "runs/sonic_unified_sft_v1/native_test.manifest.json", "episodes_sha256": "998f3d2eeebfab5e5d6415e2fb0e7b5cb36f3b66f333ff8be12276755c7a375d", "manifest_sha256": "0ddd6bb9bebdc6cf17e7aa820e88e3c6a20468387fdd3f620538d3a01bf38056", "cases_sha256": "d00f5b6aeaa9f2aac3e4b5010b31df4e86f77674cb588e17bc3f7d853e4a1f36", "continuation_policy_id": "0a96f6cada0e26a7eb1191c3428be56e63949beea528bd65334bed56121724fb", "policy": { "engine": "native_qwen_original_lm", "display_name": "Untuned Qwen", "model": "Qwen/Qwen3-0.6B", "revision": "c1899de289a04d12100db370d81485cdf75e47ca", "backend": "untrained_qwen_lm_option_logits", "checkpoint_sha256": { "model.safetensors": "f47f71177f32bcd101b7573ec9171e6a57f4f4d31148d38e382306f42996874b" }, "original_weight_files_sha256": { "model.safetensors": "f47f71177f32bcd101b7573ec9171e6a57f4f4d31148d38e382306f42996874b" }, "config_sha256": "660db3b73d788119c04535e48cf9be5f55bc3100841a718637ae695b442f27dd", "tokenizer_files_sha256": { "tokenizer.json": "aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4", "tokenizer_config.json": "d5d09f07b48c3086c508b30d1c9114bd1189145b74e982a265350c923acd8101" }, "source_sha256": { "unified_game_pipeline.py": "c176e06d7088bafbf0fbe4ca7e2be05eef6f1f5c3bf00b2d946caa21cce42313", "unified_grid_envs.py": "8ab8a9ff56f88866efc26bca41c16416e432d5ed8415cdcd33f622c5c7030663", "unified_doom_env.py": "fba56e3e97fe7bf0940962814308f63fcd2a788bf930e36f7f75b2aaa1889039", "scaled_maze.py": "4eef4a1052c913755ccccc54b9e7ee2cca5e6a1258ab748d59bfd45386d410bb", "snake_game.py": "adacc7a6013f0195e5a0952bc2a13ec14964325b52b56b2e69d780794edf549e", "evaluate_composed_maze.py": "b443469cedf81b95c154e33a7d90accd9bb7cd7423a40bfd7314be1f42fc9ae5", "predict_toy_decisions.py": "6eedc49aaf9a81763677e3cff4e87fc4e499bfc5decf0bb4678f452f578dd92b", "train_toy_decisions.py": "4f39babd5575e43d7acf3eca357329a93c60041f334bc2bd7424929bf9c0da86", "unified_jev_worker.mjs": "e844d42fcf6f1ac844c7525746cef3b8d87848682ccbaafe91e1c12294bc1dbf", "teachers.mjs": "6f8dd533e34a8f491530a26197081461e70977fce1857924ac5856c4817787ab", "evaluate_native_qwen_unified.py": "4c60f7b0ff46c8cc6b4474126ade29353ce10013857e754ad6b5518a5cdbcbab", "evaluate_native_qwen_shooting.py": "4320c9fe94ab6d04f0013e984b196c99a302c5452e887f15b988dd06b7d8e89b", "evaluate_native_qwen_navigation.py": "68aa998104645a57f4b432eaa03294da6147c58de4ed6742b3a8eb6edb4b5438", "evaluate_appo_doom.py": "d81f7329a3d7ce5e2300d98f57b0bdd876bdb0eebbe569dd99db9d363c560926", "assemble_navigation_v3_views.py": "d7256497882d17a6d799ad13feec8c018b75b5a0cbeed9fed5f3cdd5f7749d5c", "evaluate_navigation_v3.py": "5c1b413e85f4a8f43516792f003faab83e5131d29e43c8866a53c5ba52de2dc8" }, "controller": "greedy", "epsilon": 0.1, "sampling_seed": 17, "temperature": 1.0, "tie_break": "lexicographic_first", "environment_contract": "finite_task_deadline_v1", "project_training_steps": 0, "head": "unchanged original vocabulary output embeddings; no project decision head", "probability_semantics": "next-token distribution conditional on an offered A-D token; not task-success probability", "prompt": "unchanged full standard policy_request rendered by the existing native baseline prompt; official chat template, enable_thinking=False", "token_ids": { "A": 32, "B": 33, "C": 34, "D": 35 }, "precision": "bf16", "max_length": 8192, "truncation": false, "generated_tokens": 0, "forced_singletons": "Code action; no model forward; consumes one controller RNG draw" } }, "validation": { "matched_episodes": 548, "controller_transitions_checked": 19818, "physical_replay_performed_by_this_script": false }, "by_split_category": { "test/maze": { "n": 10, "successes": 2, "success_rate": 0.2, "wilson95": [ 0.05668215145437522, 0.5098375284633582 ], "mean_steps": 198.4, "mean_decisions": 163.7, "metrics": { "collisions": { "n": 10, "mean": 108.9, "min": 13.0, "max": 414.0 }, "decision_steps": { "n": 10, "mean": 198.4, "min": 27.0, "max": 752.0 }, "physical_steps": { "n": 10, "mean": 210.6, "min": 27.0, "max": 800.0 }, "remaining_steps": { "n": 10, "mean": 0.6, "min": 0.0, "max": 5.0 }, "reposition_steps": { "n": 10, "mean": 12.2, "min": 0.0, "max": 60.0 }, "visited_cells": { "n": 10, "mean": 90.5, "min": 12.0, "max": 339.0 } } }, "test/snake": { "n": 8, "successes": 0, "success_rate": 0.0, "wilson95": [ 0.0, 0.32440756488388023 ], "mean_steps": 4.875, "mean_decisions": 4.875, "metrics": { "body_length": { "n": 8, "mean": 3.25, "min": 3.0, "max": 4.0 }, "collisions": { "n": 8, "mean": 1.0, "min": 1.0, "max": 1.0 }, "decision_steps": { "n": 8, "mean": 4.875, "min": 4.0, "max": 7.0 }, "food_collected": { "n": 8, "mean": 0.25, "min": 0.0, "max": 1.0 }, "initial_score": { "n": 8, "mean": 0.0, "min": 0.0, "max": 0.0 }, "physical_steps": { "n": 8, "mean": 4.875, "min": 4.0, "max": 7.0 }, "remaining_steps": { "n": 8, "mean": 107.125, "min": 89.0, "max": 154.0 }, "score": { "n": 8, "mean": 0.25, "min": 0.0, "max": 1.0 }, "survival_steps": { "n": 8, "mean": 3.875, "min": 3.0, "max": 6.0 }, "target_food": { "n": 8, "mean": 2.25, "min": 2.0, "max": 3.0 } } }, "test/basic": { "n": 128, "successes": 56, "success_rate": 0.4375, "wilson95": [ 0.3546231312705304, 0.5240189882636581 ], "mean_steps": 47.5, "mean_decisions": 47.5, "metrics": { "ammo": { "n": 128, "mean": 36.5546875, "min": 29.0, "max": 49.0 }, "ammo_consumed": { "n": 128, "mean": 13.4453125, "min": 1.0, "max": 21.0 }, "damage_dealt": { "n": 128, "mean": 4.4140625, "min": 0.0, "max": 15.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 47.5, "min": 2.0, "max": 72.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.4375, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.4375, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": -208.7421875, "min": -391.0, "max": 96.0 }, "physical_ticks": { "n": 128, "mean": 187.890625, "min": 5.0, "max": 286.0 } } }, "test/predict_position": { "n": 128, "successes": 11, "success_rate": 0.0859375, "wilson95": [ 0.0486631313713862, 0.14734091054261297 ], "mean_steps": 27.2578125, "mean_decisions": 27.2578125, "metrics": { "ammo": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "ammo_consumed": { "n": 128, "mean": 1.0, "min": 1.0, "max": 1.0 }, "damage_dealt": { "n": 128, "mean": 16.2421875, "min": 0.0, "max": 208.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 27.2578125, "min": 10.0, "max": 71.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.0859375, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.0859375, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": -0.02204687500000007, "min": -0.2840000000000002, "max": 0.962 }, "physical_ticks": { "n": 128, "mean": 107.984375, "min": 38.0, "max": 284.0 } } }, "ood/maze": { "n": 10, "successes": 1, "success_rate": 0.1, "wilson95": [ 0.01787621309507289, 0.4041500267952385 ], "mean_steps": 268.8, "mean_decisions": 218.0, "metrics": { "collisions": { "n": 10, "mean": 144.3, "min": 76.0, "max": 416.0 }, "decision_steps": { "n": 10, "mean": 268.8, "min": 146.0, "max": 770.0 }, "physical_steps": { "n": 10, "mean": 287.4, "min": 154.0, "max": 800.0 }, "remaining_steps": { "n": 10, "mean": 0.6, "min": 0.0, "max": 6.0 }, "reposition_steps": { "n": 10, "mean": 18.6, "min": 0.0, "max": 100.0 }, "visited_cells": { "n": 10, "mean": 125.5, "min": 65.0, "max": 381.0 } } }, "ood/snake": { "n": 8, "successes": 0, "success_rate": 0.0, "wilson95": [ 0.0, 0.32440756488388023 ], "mean_steps": 5.375, "mean_decisions": 5.375, "metrics": { "body_length": { "n": 8, "mean": 3.0, "min": 3.0, "max": 3.0 }, "collisions": { "n": 8, "mean": 1.0, "min": 1.0, "max": 1.0 }, "decision_steps": { "n": 8, "mean": 5.375, "min": 5.0, "max": 7.0 }, "food_collected": { "n": 8, "mean": 0.0, "min": 0.0, "max": 0.0 }, "initial_score": { "n": 8, "mean": 0.0, "min": 0.0, "max": 0.0 }, "physical_steps": { "n": 8, "mean": 5.375, "min": 5.0, "max": 7.0 }, "remaining_steps": { "n": 8, "mean": 154.625, "min": 153.0, "max": 155.0 }, "score": { "n": 8, "mean": 0.0, "min": 0.0, "max": 0.0 }, "survival_steps": { "n": 8, "mean": 4.375, "min": 4.0, "max": 6.0 }, "target_food": { "n": 8, "mean": 3.0, "min": 3.0, "max": 3.0 } } }, "ood/basic": { "n": 128, "successes": 59, "success_rate": 0.4609375, "wilson95": [ 0.3769811559809195, 0.5471701687279482 ], "mean_steps": 24.4296875, "mean_decisions": 24.4296875, "metrics": { "ammo": { "n": 128, "mean": 36.3046875, "min": 29.0, "max": 49.0 }, "ammo_consumed": { "n": 128, "mean": 13.6953125, "min": 1.0, "max": 21.0 }, "damage_dealt": { "n": 128, "mean": 4.84375, "min": 0.0, "max": 15.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 24.4296875, "min": 1.0, "max": 36.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.4609375, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.4609375, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": -212.296875, "min": -391.0, "max": 96.0 }, "physical_ticks": { "n": 128, "mean": 192.6796875, "min": 5.0, "max": 286.0 } } }, "ood/predict_position": { "n": 128, "successes": 9, "success_rate": 0.0703125, "wilson95": [ 0.03742976830393997, 0.1282348034936063 ], "mean_steps": 18.5, "mean_decisions": 18.5, "metrics": { "ammo": { "n": 128, "mean": 0.0546875, "min": 0.0, "max": 1.0 }, "ammo_consumed": { "n": 128, "mean": 0.9453125, "min": 0.0, "max": 1.0 }, "damage_dealt": { "n": 128, "mean": 14.8046875, "min": 0.0, "max": 196.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 18.5, "min": 7.0, "max": 36.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.0703125, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.0703125, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": -0.07619531250000007, "min": -0.2840000000000002, "max": 0.948 }, "physical_ticks": { "n": 128, "mean": 146.5078125, "min": 52.0, "max": 284.0 } } } }, "macro_by_split": { "test": { "success_rate": 0.15390625, "weights": { "maze": 0.3333333333333333, "snake": 0.3333333333333333, "basic": 0.16666666666666666, "predict_position": 0.16666666666666666 }, "interval": null }, "ood": { "success_rate": 0.12187499999999998, "weights": { "maze": 0.3333333333333333, "snake": 0.3333333333333333, "basic": 0.16666666666666666, "predict_position": 0.16666666666666666 }, "interval": null } }, "paired_vs_jev": { "test/maze": { "n": 10, "wins_vs_jev": 0, "losses_vs_jev": 5, "both_success": 2, "both_failure": 3, "success_rate_delta_vs_jev": -0.5, "exact_mcnemar_two_sided_p": 0.0625, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [ "test-maze16-7830000", "test-maze8-308002", "test-maze8-308034", "test-maze8-308040", "test-maze8-308053" ] }, "test/snake": { "n": 8, "wins_vs_jev": 0, "losses_vs_jev": 8, "both_success": 0, "both_failure": 0, "success_rate_delta_vs_jev": -1.0, "exact_mcnemar_two_sided_p": 0.0078125, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [ "test-snake12-7832000", "test-snake12-7832001", "test-snake8-318000", "test-snake8-318001", "test-snake8-318002", "test-snake8-318003", "test-snake8-318004", "test-snake8-318005" ] }, "test/basic": { "n": 128, "wins_vs_jev": 0, "losses_vs_jev": 0, "both_success": 56, "both_failure": 72, "success_rate_delta_vs_jev": 0.0, "exact_mcnemar_two_sided_p": 1.0, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [] }, "test/predict_position": { "n": 128, "wins_vs_jev": 0, "losses_vs_jev": 0, "both_success": 11, "both_failure": 117, "success_rate_delta_vs_jev": 0.0, "exact_mcnemar_two_sided_p": 1.0, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [] }, "ood/maze": { "n": 10, "wins_vs_jev": 1, "losses_vs_jev": 3, "both_success": 0, "both_failure": 6, "success_rate_delta_vs_jev": -0.2, "exact_mcnemar_two_sided_p": 0.625, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [ "ood-maze8-408003" ], "loss_case_ids": [ "ood-maze16-7840001", "ood-maze8-408002", "ood-maze8-408004" ] }, "ood/snake": { "n": 8, "wins_vs_jev": 0, "losses_vs_jev": 6, "both_success": 0, "both_failure": 2, "success_rate_delta_vs_jev": -0.75, "exact_mcnemar_two_sided_p": 0.03125, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [ "ood-snake12-7842000", "ood-snake12-7842001", "ood-snake8-418000", "ood-snake8-418001", "ood-snake8-418003", "ood-snake8-418004" ] }, "ood/basic": { "n": 128, "wins_vs_jev": 0, "losses_vs_jev": 0, "both_success": 59, "both_failure": 69, "success_rate_delta_vs_jev": 0.0, "exact_mcnemar_two_sided_p": 1.0, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [], "loss_case_ids": [] }, "ood/predict_position": { "n": 128, "wins_vs_jev": 1, "losses_vs_jev": 0, "both_success": 8, "both_failure": 119, "success_rate_delta_vs_jev": 0.0078125, "exact_mcnemar_two_sided_p": 1.0, "p_adjustment": "none; descriptive planned comparisons", "win_case_ids": [ "ood-sonic_predict_position-9300827" ], "loss_case_ids": [] } }, "macro_delta_vs_jev": { "test": -0.5, "ood": -0.31536458333333334 } } }, "api_accounting": { "reported_cost_usd": "0", "unknown_reserved_usd": "0.476", "accounted_usd": "0.476", "successful_requests": 19111, "failed_requests": 236, "children": [ { "shard": 0, "manifest_sha256": "9c5987645286ca13a64a6832100f63ff73e871558dd5499863a1ae935ffda537", "episode_sha256": "9d6947068f870549cf8a0af535e6558b03b081d8b0c4120b56fb65eac1e83eb1", "cases_sha256": "6015fd00f38622df1a2c4b8b25388a77b30688d13e449f4703b3b410e7573f89", "budget_usd": "0.396000000", "accounting": { "journal_sha256": "4fe124a3facc0cab6fac1f1d16a4217214e04ad4546667b989043c6b4911566f", "journal_bytes": 21286555, "rows": 11773, "started_calls": 5887, "succeeded_calls": 5813, "failed_calls": 73, "reported_cost_usd": "0", "unknown_reserved_usd": "0.148", "accounted_usd": "0.148", "unresolved_started_ids": [ "nanojev-live-001598" ], "unknown_cost_ids": [ "nanojev-live-000148", "nanojev-live-000177", "nanojev-live-000458", "nanojev-live-000459", "nanojev-live-000481", "nanojev-live-000489", "nanojev-live-000548", "nanojev-live-000582", "nanojev-live-000618", "nanojev-live-001598", "nanojev-live-001605", "nanojev-live-001620", "nanojev-live-001660", "nanojev-live-001753", "nanojev-live-001790", "nanojev-live-001793", "nanojev-live-001794", "nanojev-live-001795", "nanojev-live-001797", "nanojev-live-001798", "nanojev-live-001799", "nanojev-live-001803", "nanojev-live-001804", "nanojev-live-001820", "nanojev-live-001853", "nanojev-live-002086", "nanojev-live-002433", "nanojev-live-002602", "nanojev-live-002604", "nanojev-live-002605", "nanojev-live-002608", "nanojev-live-002610", "nanojev-live-002635", "nanojev-live-002639", "nanojev-live-002642", "nanojev-live-002643", "nanojev-live-002664", "nanojev-live-002682", "nanojev-live-002684", "nanojev-live-002714", "nanojev-live-002716", "nanojev-live-002723", "nanojev-live-002752", "nanojev-live-002753", "nanojev-live-002764", "nanojev-live-002766", "nanojev-live-002809", "nanojev-live-002831", "nanojev-live-002857", "nanojev-live-002860", "nanojev-live-002971", "nanojev-live-002974", "nanojev-live-003142", "nanojev-live-003178", "nanojev-live-003198", "nanojev-live-003200", "nanojev-live-003218", "nanojev-live-003258", "nanojev-live-003286", "nanojev-live-003741", "nanojev-live-003767", "nanojev-live-003876", "nanojev-live-004333", "nanojev-live-004795", "nanojev-live-004838", "nanojev-live-005060", "nanojev-live-005061", "nanojev-live-005083", "nanojev-live-005466", "nanojev-live-005502", "nanojev-live-005548", "nanojev-live-005646", "nanojev-live-005739", "nanojev-live-005761" ], "zero_reported_cost_calls": 5813 } }, { "shard": 1, "manifest_sha256": "1ffb643e9c155db616ba82a1e31a8c628821a6fa5c2bff7ae063c234842bd3e5", "episode_sha256": "605e1cc4f7e1704bb9ff8b4e4824188dc3ff321c6ab7010b15465ce676a33dac", "cases_sha256": "a105c265c10d8c6e433ae9d4e633b4c35f4e634ea0fe0321b42b323918b316d1", "budget_usd": "0.376000000", "accounting": { "journal_sha256": "4b0bbb8ba3b995c6a8ca47a8e178a032216e592f3a5d6dfcac54b010d39fc85e", "journal_bytes": 6830139, "rows": 6650, "started_calls": 3325, "succeeded_calls": 3289, "failed_calls": 36, "reported_cost_usd": "0", "unknown_reserved_usd": "0.072", "accounted_usd": "0.072", "unresolved_started_ids": [], "unknown_cost_ids": [ "nanojev-live-000013", "nanojev-live-000077", "nanojev-live-000126", "nanojev-live-000150", "nanojev-live-000157", "nanojev-live-000158", "nanojev-live-000160", "nanojev-live-000161", "nanojev-live-000163", "nanojev-live-000165", "nanojev-live-000166", "nanojev-live-000169", "nanojev-live-000170", "nanojev-live-000172", "nanojev-live-001198", "nanojev-live-001224", "nanojev-live-001286", "nanojev-live-001287", "nanojev-live-001310", "nanojev-live-001314", "nanojev-live-001335", "nanojev-live-001336", "nanojev-live-001353", "nanojev-live-001374", "nanojev-live-001377", "nanojev-live-001393", "nanojev-live-001502", "nanojev-live-001654", "nanojev-live-001881", "nanojev-live-001885", "nanojev-live-001968", "nanojev-live-002000", "nanojev-live-002004", "nanojev-live-002007", "nanojev-live-002385", "nanojev-live-002799" ], "zero_reported_cost_calls": 3289 } }, { "shard": 2, "manifest_sha256": "fe050b040b7346745bf2310749bbd008274430256b3d2bb6013c268de2b8277d", "episode_sha256": "f87f93d19039bcf20fc02280555aa71b771bdf4132ba302abc28fb425be3a192", "cases_sha256": "e73536d4c33c0ed8d01f63bd269c5f982f4853afac3bc65faa63eb3a9882484f", "budget_usd": "0.376000000", "accounting": { "journal_sha256": "be2bb6b14b6d136da3a32213cc8062e784f26dd016c95d456d9d9aa38f2896ed", "journal_bytes": 4703872, "rows": 4588, "started_calls": 2294, "succeeded_calls": 2268, "failed_calls": 26, "reported_cost_usd": "0", "unknown_reserved_usd": "0.052", "accounted_usd": "0.052", "unresolved_started_ids": [], "unknown_cost_ids": [ "nanojev-live-000051", "nanojev-live-000054", "nanojev-live-000173", "nanojev-live-000196", "nanojev-live-000424", "nanojev-live-000443", "nanojev-live-000445", "nanojev-live-000446", "nanojev-live-000449", "nanojev-live-000450", "nanojev-live-000451", "nanojev-live-000453", "nanojev-live-000454", "nanojev-live-000890", "nanojev-live-000963", "nanojev-live-001021", "nanojev-live-001132", "nanojev-live-001134", "nanojev-live-001182", "nanojev-live-001192", "nanojev-live-002036", "nanojev-live-002049", "nanojev-live-002071", "nanojev-live-002223", "nanojev-live-002224", "nanojev-live-002269" ], "zero_reported_cost_calls": 2268 } }, { "shard": 3, "manifest_sha256": "6bef4ff61d752a4a0e743b717bdd1e6ee2f43f76474dd7811fe95a4cb2bffb84", "episode_sha256": "c02f4e3c4e8844124d632353e2c87907592fa348c85b7f5917c7082abce2a9b8", "cases_sha256": "a66bd9b951a99ad349ff4eb2776336faae3348925e49b6f4d161bb5909c67b4d", "budget_usd": "0.366000000", "accounting": { "journal_sha256": "0340c85535b481b542f86b42441b8aac9ef6076790361f88b5b98b135c6c1ce6", "journal_bytes": 3276727, "rows": 3207, "started_calls": 1604, "succeeded_calls": 1584, "failed_calls": 19, "reported_cost_usd": "0", "unknown_reserved_usd": "0.040", "accounted_usd": "0.040", "unresolved_started_ids": [ "nanojev-live-000146" ], "unknown_cost_ids": [ "nanojev-live-000049", "nanojev-live-000057", "nanojev-live-000105", "nanojev-live-000139", "nanojev-live-000144", "nanojev-live-000145", "nanojev-live-000146", "nanojev-live-000147", "nanojev-live-000184", "nanojev-live-000206", "nanojev-live-000217", "nanojev-live-000219", "nanojev-live-000697", "nanojev-live-000698", "nanojev-live-000700", "nanojev-live-000812", "nanojev-live-000877", "nanojev-live-000881", "nanojev-live-000967", "nanojev-live-001227" ], "zero_reported_cost_calls": 1584 } }, { "shard": 4, "manifest_sha256": "fa50bcdba7680212c61a486d6777330e59a0a22f1aac23c9f046df44ded40654", "episode_sha256": "dadce12f5006d8da6eaf68ac4c3a24b8f4ec0074b4490bf5e510d55c3d54eeea", "cases_sha256": "1daad9c6f3aa68d1473e58de77b1c1ef42ba719f3ac305fb5e7595ae188339e9", "budget_usd": "0.378000000", "accounting": { "journal_sha256": "84941c60b35c39a04bcb732a15e0413fd4ba429c9b1d387308fe14cabdb21a30", "journal_bytes": 4901637, "rows": 3632, "started_calls": 1816, "succeeded_calls": 1791, "failed_calls": 25, "reported_cost_usd": "0", "unknown_reserved_usd": "0.050", "accounted_usd": "0.050", "unresolved_started_ids": [], "unknown_cost_ids": [ "nanojev-live-000051", "nanojev-live-000071", "nanojev-live-000081", "nanojev-live-000113", "nanojev-live-000155", "nanojev-live-000156", "nanojev-live-000157", "nanojev-live-000158", "nanojev-live-000159", "nanojev-live-000161", "nanojev-live-000162", "nanojev-live-000167", "nanojev-live-000168", "nanojev-live-000169", "nanojev-live-000579", "nanojev-live-000790", "nanojev-live-001146", "nanojev-live-001186", "nanojev-live-001385", "nanojev-live-001422", "nanojev-live-001735", "nanojev-live-001758", "nanojev-live-001768", "nanojev-live-001789", "nanojev-live-001797" ], "zero_reported_cost_calls": 1791 } }, { "shard": 5, "manifest_sha256": "0babfc2b9be9781df93910af1b12aabf7e19fc82ce077c976c733c372db3095a", "episode_sha256": "70227824646ecd0380833dede8d84bb68b9651949e4b3c1656aa033609655466", "cases_sha256": "f0c9774b0f9488f5d493107bcd72e691b551cc273e0ae9b14809723b096b2a27", "budget_usd": "0.362000000", "accounting": { "journal_sha256": "5e89024f07ef5402d6a9c71fb3796ed256746441081ccbcf07bce8535bd5b2e2", "journal_bytes": 5380219, "rows": 3714, "started_calls": 1857, "succeeded_calls": 1841, "failed_calls": 16, "reported_cost_usd": "0", "unknown_reserved_usd": "0.032", "accounted_usd": "0.032", "unresolved_started_ids": [], "unknown_cost_ids": [ "nanojev-live-000021", "nanojev-live-000118", "nanojev-live-000174", "nanojev-live-000182", "nanojev-live-000191", "nanojev-live-000197", "nanojev-live-000260", "nanojev-live-000401", "nanojev-live-000409", "nanojev-live-000607", "nanojev-live-000956", "nanojev-live-001070", "nanojev-live-001071", "nanojev-live-001120", "nanojev-live-001575", "nanojev-live-001671" ], "zero_reported_cost_calls": 1841 } }, { "shard": 6, "manifest_sha256": "94fcec5755d6684b91f2d63b53c183abcc95d8276fab46f46df6d8c3d8a5cc17", "episode_sha256": "8836cb7138b652955b9c847e68a25d6251a2038f87299866b1d0d5455c30d6fe", "cases_sha256": "6c49eef6a2bb9d5cc28f3c44fa46123fb2beef979bd56d156ac7e09cdeb28c65", "budget_usd": "0.366000000", "accounting": { "journal_sha256": "5462f2367943c2bcbb5dffc1e966afb0f92ca00730f465f697ecbe834b9373e1", "journal_bytes": 3508522, "rows": 2634, "started_calls": 1317, "succeeded_calls": 1308, "failed_calls": 9, "reported_cost_usd": "0", "unknown_reserved_usd": "0.018", "accounted_usd": "0.018", "unresolved_started_ids": [], "unknown_cost_ids": [ "nanojev-live-000020", "nanojev-live-000042", "nanojev-live-000046", "nanojev-live-000204", "nanojev-live-000338", "nanojev-live-000376", "nanojev-live-000377", "nanojev-live-000378", "nanojev-live-000429" ], "zero_reported_cost_calls": 1308 } }, { "shard": 7, "manifest_sha256": "a709a7391bccc06e1bd69129e257a35c91e2a6e71ed9b2d3665a0fad88b5dad6", "episode_sha256": "3f1d22e04720ff1ec9c0ea36df40b912529787bc86070104639f64f3b9b71c53", "cases_sha256": "6e73ff433d37cdc5f54f48e3423eac83e252f8160f1fecbaae34b472cb9ff75d", "budget_usd": "0.380000000", "accounting": { "journal_sha256": "7d762399a9dc60f62298f1cc779b1fa59eaa17eac00019b29b7dd66fa7a1aa6f", "journal_bytes": 3223202, "rows": 2498, "started_calls": 1249, "succeeded_calls": 1217, "failed_calls": 32, "reported_cost_usd": "0", "unknown_reserved_usd": "0.064", "accounted_usd": "0.064", "unresolved_started_ids": [], "unknown_cost_ids": [ "nanojev-live-000085", "nanojev-live-000086", "nanojev-live-000140", "nanojev-live-000174", "nanojev-live-000379", "nanojev-live-000380", "nanojev-live-000381", "nanojev-live-000382", "nanojev-live-000383", "nanojev-live-000384", "nanojev-live-000385", "nanojev-live-000386", "nanojev-live-000388", "nanojev-live-000389", "nanojev-live-000390", "nanojev-live-000466", "nanojev-live-000566", "nanojev-live-000576", "nanojev-live-000577", "nanojev-live-000578", "nanojev-live-000596", "nanojev-live-000599", "nanojev-live-000612", "nanojev-live-000613", "nanojev-live-000643", "nanojev-live-000669", "nanojev-live-000689", "nanojev-live-000739", "nanojev-live-000800", "nanojev-live-000813", "nanojev-live-000818", "nanojev-live-000821" ], "zero_reported_cost_calls": 1217 } } ], "receipt_identity": "(journal shard, call ID)", "scope": "Includes prior journal exactly once in shard zero; zero reported cost is not independent billing verification.", "verification": "Recomputed from the completed parallel manifest; not a live gateway balance or independent billing audit." }, "expert_reference": { "source": { "episodes_path": "runs/sonic_unified_sft_v1/expert/episodes.jsonl", "manifest_path": "runs/sonic_unified_sft_v1/expert/episodes.manifest.json", "episodes_sha256": "495a1294d47ab422c6062fa83df0b3ac4bc094eebd5798c8aa5cb367e2825f20", "manifest_sha256": "06cd5e4b9257e79bccd6152515319f71b8cf27bb5318c90b53384eb09f3b8d80", "cases_sha256": "20f7c61fcfab45d1e3e93f7d23dd6c9dda07e48b5a003f55a90c565675b58727", "continuation_policy_id": "12aa0296306df4feafe25394ec070720a9da52adb612d2060d43ef3cc7dcad35", "policy": { "engine": "sonic_doom_pure_vision_pp", "render_adapter": "old_materials_tick_mirror", "controller": "greedy", "epsilon": 0.0, "temperature": 1.0, "sampling_seed": 17, "tie_break": "lexicographic_first", "model": { "engine": "sonic_doom_pure_vision_pp", "checkpoint_sha256": "59756ba40d0e98140893548a576e1a2d024cdc201624a7964f9aca8a8ec98e37", "config_sha256": "01b09781bf43bdb81fdc1ba6335e36f72841019612821769f0e865b5a7636055", "source_commit": "392f036b62e48f4be4394651a646182cb91a14ff", "adapter_sha256": "01afba8767bbafae68109cd3add479f829d2b443c26eabdf926c8498ee16c7a2", "weights_only": true, "numeric_allowlist": [ "numpy.core.multiarray.scalar", "numpy.dtype", "numpy.float64_dtype" ], "strict_state_load": true, "state_tensors": 25, "parameter_count": 2871525, "train_step": 97758, "env_steps": 400416768, "reported_training_reward": 0.9201099999999998, "action_order": [ "noop", "left", "right", "shoot" ], "device": "cpu", "precision": "float32", "rnn": "GRU 512, zero state per episode", "input": "uint8 RGB HWC 120x160 -> nearest 72x128 -> CHW -> /255 -> saved RMS", "sound_encoder": false, "aim_assistance": false, "raw_state_tensor_sha256": { "obs_normalizer.running_mean_std.running_mean_std.img.running_mean": "2b7dfc3215ae632ee1b10cb03b0228087b16feef0648108222e31b594d78ea95", "obs_normalizer.running_mean_std.running_mean_std.img.running_var": "a1adb0aadb63b66161a1662940a8fd9288a3d6ee257ee3834dfe7e34621a2b18", "obs_normalizer.running_mean_std.running_mean_std.img.count": "463f764e9f83d00e606d7c7f4a42fc02f3dc76b607844e59436071a0c56db1b5", "obs_normalizer.running_mean_std.running_mean_std.sound.running_mean": "0b1b29ef265c6d9b17074289e7acf02b1ecaef4957119287d2c444e910ec4bb8", "obs_normalizer.running_mean_std.running_mean_std.sound.running_var": "3436351cdb420b9352bd630351301716b9927f35b3071446f53d6f50d28b357f", "obs_normalizer.running_mean_std.running_mean_std.sound.count": "463f764e9f83d00e606d7c7f4a42fc02f3dc76b607844e59436071a0c56db1b5", "returns_normalizer.running_mean": "dacc9980c7edb99cb644fc2fe9e4bb1df2a4fae953395b0a488557e946fdb379", "returns_normalizer.running_var": "ffcedc6a3ed829c1efd4d2e53c0d156fdef205628bf565148579993378e26092", "returns_normalizer.count": "defe8b5551f5cbd9d19db27c85810a5b0f743aa402d83faddbab3ff5d9d01c83", "encoder.basic_encoder.enc.conv_head.0.weight": "9616916dbb0efa62f40a52e990f4989d260e1958536dd0b84de6d3747fe45176", "encoder.basic_encoder.enc.conv_head.0.bias": "feffc9654379fe9499827e4c915fe666b82baf94494edcd0cd68cbded16961e3", "encoder.basic_encoder.enc.conv_head.2.weight": "5946693b13ff37568b244f081b9ddf82db18cd99d8b0c2abd7475726b8ec8321", "encoder.basic_encoder.enc.conv_head.2.bias": "c01a24348b80fc7325dca9535a3241885cf28745edd9a563e517083e1c7ca36e", "encoder.basic_encoder.enc.conv_head.4.weight": "324ed34f7c12508f9ff700b731b8482b0e73a09f0265b7d0b6ed0d173496be17", "encoder.basic_encoder.enc.conv_head.4.bias": "ea1ce8683a57c6f386dceccd38979d7fad7aaae9e665d847019d5edf8e57f3ec", "encoder.basic_encoder.enc.mlp_layers.0.weight": "2fb242b17db53005243a994dbbcdd30ee231b69dc96192bf500868e9cf60fcd1", "encoder.basic_encoder.enc.mlp_layers.0.bias": "f778dc4a779a944fc8038e9f7b35d16ddeeff2043bd69923670406e5f88fbe0e", "core.core.weight_ih_l0": "9b0836b6e7d5e14f8fcd64242037a0e55ebe329506be2d3ea73c470afa814800", "core.core.weight_hh_l0": "3b13c6c19bf49b849b940617012bd52bb9da0a1d2e4b56cc5aaaa5a6473a46fd", "core.core.bias_ih_l0": "2b9b99149f1fe625d56bcf56b7ae6ba466ba4dca269230648ce2aab409aa77f1", "core.core.bias_hh_l0": "9c1adb72c2737a1dacf8a9d30c7857399e850c1ecc54a58da98d6a9887af99ea", "critic_linear.weight": "0650b06aedb84dcdfe812bc991b0bda5f723c0203801dc1ff6900ce7e9cf7b13", "critic_linear.bias": "e247f4f7525622fd328033336d74f8ab69c58107f768cfce3251f6a826a1e03c", "action_parameterization.distribution_linear.weight": "55cf6e6dcb693a751efaf7c507fb3580371dcebb31fb2ce44b315e41014a8115", "action_parameterization.distribution_linear.bias": "a96c628e91f765361dd466eb685d5d7e7a19981f695892f0f3a8c91db8bdc661" }, "model": "thainv0212/sonic_doom", "revision": "392f036b62e48f4be4394651a646182cb91a14ff", "cfg_sha256": "01b09781bf43bdb81fdc1ba6335e36f72841019612821769f0e865b5a7636055" }, "checkpoint_sha256": { "model": "59756ba40d0e98140893548a576e1a2d024cdc201624a7964f9aca8a8ec98e37", "cfg": "01b09781bf43bdb81fdc1ba6335e36f72841019612821769f0e865b5a7636055" }, "source_sha256": { "sonic_predict_data.py": "9a0b0d40d505f21f93f698a24bc14bccd7ea5c2b17a63868b78cfd2adbd770a5", "sonic_predict_policy.py": "01afba8767bbafae68109cd3add479f829d2b443c26eabdf926c8498ee16c7a2", "appo_basic_data.py": "5ff42cbd9090f9a91500a99f6a479b6736f25a299cefcfbb4110cc7923a57cbc", "evaluate_appo_doom.py": "d81f7329a3d7ce5e2300d98f57b0bdd876bdb0eebbe569dd99db9d363c560926", "unified_doom_env.py": "fba56e3e97fe7bf0940962814308f63fcd2a788bf930e36f7f75b2aaa1889039", "unified_game_pipeline.py": "c176e06d7088bafbf0fbe4ca7e2be05eef6f1f5c3bf00b2d946caa21cce42313" }, "environment_contract": "finite_task_deadline_v1", "observation": "Unmodified UnifiedDoomEnv 320x240 visible-label/player structured state with history_length=4 and installed new Freedoom materials. Expert pixels, hidden state and actions are never inserted into the student observation.", "assets": { "expert_game_wad_sha256": "729f100448d0b48e12ecc8004e096a9ea6df024467d962f002bb286805be3a8e", "student_game_wad_sha256": "a8772e088847032510d97ba2312406a6998f21cbab44d4ff10696faa9c0ecd4b", "scenario_wad_sha256": "a5a1bdbbfdbef2a505b496f6889d06751ba3431a50ec382fb303fcc4066917a6", "scenario": "Both environments load the identical installed ViZDoom 1.3.0 predict_position.cfg and scenario WAD; standard reward and terminal rules apply to both." } } }, "episodes": 896, "by_split": { "train": { "n": 512, "successes": 489, "success_rate": 0.955078125, "wilson95": [ 0.9334960572612445, 0.9698822812409755 ], "mean_steps": 20.796875, "mean_decisions": 20.796875, "metrics": { "ammo": { "n": 512, "mean": 0.0, "min": 0.0, "max": 0.0 }, "ammo_consumed": { "n": 512, "mean": 1.0, "min": 1.0, "max": 1.0 }, "damage_dealt": { "n": 512, "mean": 153.16015625, "min": 0.0, "max": 228.0 }, "damage_taken": { "n": 512, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 512, "mean": 20.796875, "min": 10.0, "max": 65.0 }, "health": { "n": 512, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 512, "mean": 0.955078125, "min": 0.0, "max": 1.0 }, "kills": { "n": 512, "mean": 0.955078125, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 512, "mean": 0.873333984375, "min": -0.2590000000000002, "max": 0.963 }, "physical_ticks": { "n": 512, "mean": 81.744140625, "min": 37.0, "max": 259.0 } } }, "dev": { "n": 64, "successes": 62, "success_rate": 0.96875, "wilson95": [ 0.8930270822904168, 0.9913878616538282 ], "mean_steps": 22.390625, "mean_decisions": 22.390625, "metrics": { "ammo": { "n": 64, "mean": 0.0, "min": 0.0, "max": 0.0 }, "ammo_consumed": { "n": 64, "mean": 1.0, "min": 1.0, "max": 1.0 }, "damage_dealt": { "n": 64, "mean": 153.71875, "min": 0.0, "max": 216.0 }, "damage_taken": { "n": 64, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 64, "mean": 22.390625, "min": 10.0, "max": 56.0 }, "health": { "n": 64, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 64, "mean": 0.96875, "min": 0.0, "max": 1.0 }, "kills": { "n": 64, "mean": 0.96875, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 64, "mean": 0.8807031249999999, "min": -0.07500000000000004, "max": 0.962 }, "physical_ticks": { "n": 64, "mean": 88.046875, "min": 38.0, "max": 223.0 } } }, "calibration": { "n": 64, "successes": 58, "success_rate": 0.90625, "wilson95": [ 0.810171204003544, 0.9563217474148017 ], "mean_steps": 23.4375, "mean_decisions": 23.4375, "metrics": { "ammo": { "n": 64, "mean": 0.0, "min": 0.0, "max": 0.0 }, "ammo_consumed": { "n": 64, "mean": 1.0, "min": 1.0, "max": 1.0 }, "damage_dealt": { "n": 64, "mean": 147.96875, "min": 0.0, "max": 216.0 }, "damage_taken": { "n": 64, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 64, "mean": 23.4375, "min": 10.0, "max": 51.0 }, "health": { "n": 64, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 64, "mean": 0.90625, "min": 0.0, "max": 1.0 }, "kills": { "n": 64, "mean": 0.90625, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 64, "mean": 0.8139687499999999, "min": -0.17900000000000013, "max": 0.961 }, "physical_ticks": { "n": 64, "mean": 92.28125, "min": 39.0, "max": 202.0 } } }, "test": { "n": 128, "successes": 125, "success_rate": 0.9765625, "wilson95": [ 0.9333562549485678, 0.9919975836032445 ], "mean_steps": 18.5390625, "mean_decisions": 18.5390625, "metrics": { "ammo": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "ammo_consumed": { "n": 128, "mean": 1.0, "min": 1.0, "max": 1.0 }, "damage_dealt": { "n": 128, "mean": 153.4453125, "min": 0.0, "max": 216.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 18.5390625, "min": 10.0, "max": 51.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.9765625, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.9765625, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": 0.90403125, "min": -0.16300000000000012, "max": 0.962 }, "physical_ticks": { "n": 128, "mean": 72.53125, "min": 38.0, "max": 202.0 } } }, "ood": { "n": 128, "successes": 61, "success_rate": 0.4765625, "wilson95": [ 0.3919887260658145, 0.5625020687595063 ], "mean_steps": 12.0625, "mean_decisions": 12.0625, "metrics": { "ammo": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "ammo_consumed": { "n": 128, "mean": 1.0, "min": 1.0, "max": 1.0 }, "damage_dealt": { "n": 128, "mean": 92.6640625, "min": 0.0, "max": 223.0 }, "damage_taken": { "n": 128, "mean": 0.0, "min": 0.0, "max": 0.0 }, "decisions": { "n": 128, "mean": 12.0625, "min": 6.0, "max": 25.0 }, "health": { "n": 128, "mean": 100.0, "min": 100.0, "max": 100.0 }, "killcount": { "n": 128, "mean": 0.4765625, "min": 0.0, "max": 1.0 }, "kills": { "n": 128, "mean": 0.4765625, "min": 0.0, "max": 1.0 }, "native_reward": { "n": 128, "mean": 0.38218749999999996, "min": -0.19100000000000011, "max": 0.957 }, "physical_ticks": { "n": 128, "mean": 94.375, "min": 43.0, "max": 198.0 } } } }, "comparison_scope": "PP RGB expert collection, greedy epsilon=0. Uses old-material synchronized rendering and recurrent pixel inputs. Not a same-observation or same-controller primary comparison." }, "notes": [ "All primary runs cover the exact same complete frozen cases; no result-dependent filtering.", "Macro weights are Maze 1/3, Snake 1/3, Basic 1/6 and PP 1/6; no 548-case micro-average headline.", "Wilson intervals describe episode-level rates; exact McNemar tests use paired episode outcomes.", "Expert-action fit is not eventual-success calibration. No probability-calibration claim is made here.", "Controller and recorded physical counters are verified; independent simulator replay is a separate artifact." ] }