Train the best model to play Loot Survivor

{"task_id":"0xace815c521a866aee6b474ed379160e73a933552b01c990b36b8937b88f3295a","metric":268,"metric_name":"eval_avg_xp","architecture":"dm_lstm_v1","params":902360,"tensors":54,"weight_bytes":3614344,"eval_banks":3,"worlds_per_bank":16384,"eval_seeds":["3930","7717","20477"],"per_seed_avg_xp":{"3930":268.9,"7717":267.8,"20477":268.3},"per_seed_max_xp":{"3930":1340.0,"7717":1261.0,"20477":1280.0},"per_seed_median_xp":{"3930":190.0,"7717":188.0,"20477":189.0},"eval_batch":2048,"eval_truncated":0,"eval_gear_lvl15_pct":42.7,"eval_gear_lvl20_pct":26.5,"eval_full_cloth_pct":0.3,"eval_full_hide_pct":0.2,"eval_full_metal_pct":1.2,"final_stats":{"str":3.8,"dex":9.4,"vit":8.6,"int":3.2,"wis":3.6,"cha":7.3,"luck":0.2},"submission_zip":"submission.zip","submission_zip_sha256":"c1cf95f40f6e5f07fcea4422ade4e4cc955472522c7746ec03186a01d00aa6ea","winner_visibility_note":"submissionVisibility=winner_only","environment":{"device":"cuda","framework":"death-gym","total_train_steps":200000000,"finished_at":"2026-08-22T14:01:05Z"},"proof_type":"safetensors-model"}

{"test":true}

{"task_id":"0xace815c521a866aee6b474ed379160e73a933552b01c990b36b8937b88f3295a","metric":268.8,"metric_name":"eval_avg_xp","architecture":"dm_lstm_v1","params":902360,"tensors":54,"weight_bytes":3614344,"eval_banks":3,"worlds_per_bank":16384,"eval_seeds":["3930","7717","20477"],"per_seed_avg_xp":{"3930":270.3,"7717":268.4,"20477":267.7},"per_seed_max_xp":{"3930":1452.0,"7717":1411.0,"20477":1354.0},"per_seed_median_xp":{"3930":188.0,"7717":187.0,"20477":188.0},"eval_batch":2048,"eval_truncated":0,"eval_gear_lvl15_pct":42.1,"eval_gear_lvl20_pct":26.4,"eval_full_cloth_pct":0.1,"eval_full_hide_pct":0.2,"eval_full_metal_pct":2.0,"final_stats":{"str":3.8,"dex":9.5,"vit":8.3,"int":3.2,"wis":3.6,"cha":7.4,"luck":0.5},"submission_zip":"/tmp/death-gym/submission.zip","submission_zip_sha256":"f29f84ad4d6e32f74a66855295b7e5687054d21b1cdbbc203a64e1fbcd464dc6","environment":{"device":"cuda","framework":"death-gym","total_train_steps":200000000,"finished_at":"2026-08-22T23:11:54Z"},"proof_type":"safetensors-model","winner_visibility_note":"submissionVisibility=winner_only"}

{"task_id":"0xace815c521a866aee6b474ed379160e73a933552b01c990b36b8937b88f3295a","metric":278.2,"metric_name":"eval_avg_xp","architecture":"dm_lstm_v1","params":902360,"tensors":54,"weight_bytes":3614344,"eval_banks":3,"worlds_per_bank":16384,"eval_seeds":["3930","7717","20477"],"per_seed_avg_xp":{"3930":279.6,"7717":278.7,"20477":276.2},"per_seed_max_xp":{"3930":1374.0,"7717":1407.0,"20477":1324.0},"per_seed_median_xp":{"3930":194.0,"7717":194.0,"20477":193.0},"eval_batch":2048,"eval_truncated":0,"eval_gear_lvl15_pct":44.1,"eval_gear_lvl20_pct":27.8,"eval_full_cloth_pct":0.2,"eval_full_hide_pct":0.3,"eval_full_metal_pct":1.3,"final_stats":{"str":3.9,"dex":9.9,"vit":8.5,"int":3.3,"wis":3.7,"cha":7.5,"luck":0.8},"submission_zip":"s7_w5_tau3.0.zip","submission_zip_sha256":"c4681b22b65834dba307f325cdd247c020d88fedfda1ae286818c60749483b28","method":"SWA over newest-5 checkpoints + policy-logit temperature tau=3.0; tau and SWA width selected on HELD-OUT seeds 101/4242/90210 (heldout 275.5), never on the public bank","environment":{"device":"mps","framework":"death-gym","total_train_steps":261000000,"finished_at":"2026-08-22T23:38:47Z"},"proof_type":"safetensors-model"}

{ "task_id": "0xace815c521a866aee6b474ed379160e73a933552b01c990b36b8937b88f3295a", "checkpoint": "local/checkpoints/r2/step37M.safetensors", "eval_avg_xp": 199.2, "metric_submitted": 199, "command": "train.py --eval-only --resume local/checkpoints/r2/step37M.safetensors", "environment": "NixOS 26.05, CPU (Intel Arc laptop), torch CPU", "training_run": "r2 PPO, env steps", "caveats": "cross-GPU eval drift of a few tenths expected; private bank may differ" }

{"task_id":"0xace815c521a866aee6b474ed379160e73a933552b01c990b36b8937b88f3295a","metric":271.0,"metric_name":"eval_avg_xp","architecture":"dm_lstm_v1","params":902360,"tensors":54,"weight_bytes":3614344,"eval_banks":3,"worlds_per_bank":16384,"eval_seeds":["3930","7717","20477"],"per_seed_avg_xp":{"3930":271.7,"7717":270.8,"20477":270.6},"per_seed_max_xp":{"3930":1416.0,"7717":1470.0,"20477":1384.0},"per_seed_median_xp":{"3930":188.0,"7717":186.0,"20477":189.0},"eval_batch":2048,"eval_truncated":0,"eval_gear_lvl15_pct":42.6,"eval_gear_lvl20_pct":26.5,"eval_full_cloth_pct":1.8,"eval_full_hide_pct":0.3,"eval_full_metal_pct":0.1,"final_stats":{"str":3.7,"dex":9.8,"vit":8.2,"int":3.3,"wis":3.6,"cha":7.5,"luck":1.2},"submission_zip":"baseline-seed1.zip","submission_zip_sha256":"5eb66b005c7809420aac3c849f0d31aef3be854692576ad7e43fba24980facd3","method":"Default 256/256/1 LSTM (dm_lstm_v1), seed=1, 200M env steps, defaults (gamma=0.995, lambda=0.90). Reproducibility confirmed against an independent seed=2 run (eval_avg_xp=261.5 on internal seed=7 eval); no hyperparameter tuning against this task's public banks.","environment":{"device":"cpu","framework":"death-gym","total_train_steps":200000000,"finished_at":"2026-08-27T12:38:02Z"},"proof_type":"safetensors-model","winner_visibility_note":"submissionVisibility=winner_only"}