| {"step": 0, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.8160546875, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 83564, "mean_completion_length": 1671.28, "median_completion_length": 1812, "min_completion_length": 992, "max_completion_length": 2048, "frac_has_think": 0.7, "frac_has_boxed": 0.7, "mean_reward_std_per_prompt": 0.2, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": null, "avg_step_time_sec": null, "gpu_allocated_gb": null, "gpu_reserved_gb": null, "gpu_peak_gb": null, "elapsed_sec": null, "eta_sec": null, "timestamp": "2026-02-15T14:29:51.533176"} |
| {"step": 1, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.22, "mean_correctness_reward": 0.22, "mean_length_penalty": -0.94798828125, "mean_total_reward": 0.22, "std_total_reward": 0.4142463035441596, "total_tokens": 97074, "mean_completion_length": 1941.48, "median_completion_length": 2048, "min_completion_length": 1079, "max_completion_length": 2048, "frac_has_think": 0.46, "frac_has_boxed": 0.46, "mean_reward_std_per_prompt": 0.2559591794226542, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 27.94786262512207, "avg_step_time_sec": 27.94786262512207, "gpu_allocated_gb": 32.365621248, "gpu_reserved_gb": 79.92246272, "gpu_peak_gb": 73.906566656, "elapsed_sec": 27.952669382095337, "eta_sec": 5561.624662399292, "timestamp": "2026-02-15T14:30:21.149910"} |
| {"step": 2, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.852724609375, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 87319, "mean_completion_length": 1746.38, "median_completion_length": 1950, "min_completion_length": 760, "max_completion_length": 2048, "frac_has_think": 0.74, "frac_has_boxed": 0.74, "mean_reward_std_per_prompt": 0.36963110361044393, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 29.601027250289917, "avg_step_time_sec": 28.774444937705994, "gpu_allocated_gb": 32.365654016, "gpu_reserved_gb": 79.95392, "gpu_peak_gb": 74.127090176, "elapsed_sec": 57.56005048751831, "eta_sec": 5697.340097665787, "timestamp": "2026-02-15T14:30:49.496231"} |
| {"step": 3, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.722470703125, "mean_total_reward": 0.74, "std_total_reward": 0.4386342439892262, "total_tokens": 73981, "mean_completion_length": 1479.62, "median_completion_length": 1446, "min_completion_length": 692, "max_completion_length": 2048, "frac_has_think": 0.76, "frac_has_boxed": 0.78, "mean_reward_std_per_prompt": 0.25165151389911683, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 28.27762770652771, "avg_step_time_sec": 28.6088391939799, "gpu_allocated_gb": 32.365626368, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.127090176, "elapsed_sec": 85.84399461746216, "eta_sec": 5635.941321214041, "timestamp": "2026-02-15T14:31:17.114370"} |
| {"step": 4, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.18, "mean_correctness_reward": 0.18, "mean_length_penalty": -0.944658203125, "mean_total_reward": 0.18, "std_total_reward": 0.3841874542459709, "total_tokens": 96733, "mean_completion_length": 1934.66, "median_completion_length": 2048, "min_completion_length": 1111, "max_completion_length": 2048, "frac_has_think": 0.26, "frac_has_boxed": 0.26, "mean_reward_std_per_prompt": 0.2516515138991168, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 27.60600233078003, "avg_step_time_sec": 28.35812997817993, "gpu_allocated_gb": 32.365638656, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.127090176, "elapsed_sec": 113.4564573764801, "eta_sec": 5558.193475723267, "timestamp": "2026-02-15T14:31:46.834161"} |
| {"step": 5, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.5, "mean_correctness_reward": 0.5, "mean_length_penalty": -0.827548828125, "mean_total_reward": 0.5, "std_total_reward": 0.5, "total_tokens": 84741, "mean_completion_length": 1694.82, "median_completion_length": 2048, "min_completion_length": 815, "max_completion_length": 2048, "frac_has_think": 0.62, "frac_has_boxed": 0.62, "mean_reward_std_per_prompt": 0.26, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 29.881340265274048, "avg_step_time_sec": 28.662772035598756, "gpu_allocated_gb": 32.365667328, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.156304384, "elapsed_sec": 143.34542655944824, "eta_sec": 5589.240546941757, "timestamp": "2026-02-15T14:32:15.051182"} |
| {"step": 6, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.7, "mean_correctness_reward": 0.7, "mean_length_penalty": -0.7873828125, "mean_total_reward": 0.7, "std_total_reward": 0.458257569495584, "total_tokens": 80628, "mean_completion_length": 1612.56, "median_completion_length": 1622, "min_completion_length": 736, "max_completion_length": 2048, "frac_has_think": 0.78, "frac_has_boxed": 0.78, "mean_reward_std_per_prompt": 0.2696311036104439, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 28.03603458404541, "avg_step_time_sec": 28.5583157936732, "gpu_allocated_gb": 32.365632512, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.156304384, "elapsed_sec": 171.38831543922424, "eta_sec": 5540.313263972601, "timestamp": "2026-02-15T14:32:42.780885"} |
| {"step": 7, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.46, "mean_correctness_reward": 0.46, "mean_length_penalty": -0.874619140625, "mean_total_reward": 0.46, "std_total_reward": 0.4983974317750845, "total_tokens": 89561, "mean_completion_length": 1791.22, "median_completion_length": 2048, "min_completion_length": 731, "max_completion_length": 2048, "frac_has_think": 0.54, "frac_has_boxed": 0.54, "mean_reward_std_per_prompt": 0.23797958971132713, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 28.157071828842163, "avg_step_time_sec": 28.500995227268763, "gpu_allocated_gb": 32.36562432, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.156304384, "elapsed_sec": 199.55165338516235, "eta_sec": 5500.692078862871, "timestamp": "2026-02-15T14:33:11.640562"} |
| {"step": 8, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.52, "mean_correctness_reward": 0.52, "mean_length_penalty": -0.80248046875, "mean_total_reward": 0.52, "std_total_reward": 0.49959983987187184, "total_tokens": 82174, "mean_completion_length": 1643.48, "median_completion_length": 1676, "min_completion_length": 814, "max_completion_length": 2048, "frac_has_think": 0.78, "frac_has_boxed": 0.78, "mean_reward_std_per_prompt": 0.24, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.315357208251953, "avg_step_time_sec": 28.477790474891663, "gpu_allocated_gb": 32.365616128, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.156304384, "elapsed_sec": 227.87330174446106, "eta_sec": 5467.735771179199, "timestamp": "2026-02-15T14:33:39.471177"} |
| {"step": 9, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.68412109375, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 70054, "mean_completion_length": 1401.08, "median_completion_length": 1359, "min_completion_length": 726, "max_completion_length": 2048, "frac_has_think": 0.84, "frac_has_boxed": 0.84, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.41650915145874, "avg_step_time_sec": 28.47098143895467, "gpu_allocated_gb": 32.365641728, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.156304384, "elapsed_sec": 256.298024892807, "eta_sec": 5437.957454840342, "timestamp": "2026-02-15T14:34:06.941824"} |
| {"step": 10, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.46, "mean_correctness_reward": 0.46, "mean_length_penalty": -0.932568359375, "mean_total_reward": 0.46, "std_total_reward": 0.4983974317750845, "total_tokens": 95495, "mean_completion_length": 1909.9, "median_completion_length": 2048, "min_completion_length": 1076, "max_completion_length": 2048, "frac_has_think": 0.62, "frac_has_boxed": 0.62, "mean_reward_std_per_prompt": 0.4709137211200046, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 27.307122230529785, "avg_step_time_sec": 28.354595518112184, "gpu_allocated_gb": 32.365712384, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 283.6122679710388, "eta_sec": 5387.373148441315, "timestamp": "2026-02-15T14:34:36.300324"} |
| {"step": 11, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.62, "mean_correctness_reward": 0.62, "mean_length_penalty": -0.847060546875, "mean_total_reward": 0.62, "std_total_reward": 0.4853864439804639, "total_tokens": 86739, "mean_completion_length": 1734.78, "median_completion_length": 1929, "min_completion_length": 896, "max_completion_length": 2048, "frac_has_think": 0.64, "frac_has_boxed": 0.64, "mean_reward_std_per_prompt": 0.3, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 29.035467863082886, "avg_step_time_sec": 28.41649300401861, "gpu_allocated_gb": 32.36563456, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 312.6544551849365, "eta_sec": 5370.717177759517, "timestamp": "2026-02-15T14:35:05.117724"} |
| {"step": 12, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.58, "mean_correctness_reward": 0.58, "mean_length_penalty": -0.80970703125, "mean_total_reward": 0.58, "std_total_reward": 0.4935585071701226, "total_tokens": 82914, "mean_completion_length": 1658.28, "median_completion_length": 1705, "min_completion_length": 764, "max_completion_length": 2048, "frac_has_think": 0.76, "frac_has_boxed": 0.76, "mean_reward_std_per_prompt": 0.22, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.814789056777954, "avg_step_time_sec": 28.449684341748554, "gpu_allocated_gb": 32.365643776, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 341.476256608963, "eta_sec": 5348.540656248728, "timestamp": "2026-02-15T14:35:33.091205"} |
| {"step": 13, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.64, "mean_correctness_reward": 0.64, "mean_length_penalty": -0.842578125, "mean_total_reward": 0.64, "std_total_reward": 0.48, "total_tokens": 86280, "mean_completion_length": 1725.6, "median_completion_length": 1900, "min_completion_length": 982, "max_completion_length": 2048, "frac_has_think": 0.7, "frac_has_boxed": 0.7, "mean_reward_std_per_prompt": 0.31165151389911677, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 27.95727849006653, "avg_step_time_sec": 28.411806968542244, "gpu_allocated_gb": 32.365641728, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 369.440274477005, "eta_sec": 5313.0079031174, "timestamp": "2026-02-15T14:36:01.815624"} |
| {"step": 14, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.44, "mean_correctness_reward": 0.44, "mean_length_penalty": -0.787900390625, "mean_total_reward": 0.44, "std_total_reward": 0.4963869458396343, "total_tokens": 80681, "mean_completion_length": 1613.62, "median_completion_length": 1717, "min_completion_length": 758, "max_completion_length": 2048, "frac_has_think": 0.76, "frac_has_boxed": 0.76, "mean_reward_std_per_prompt": 0.31165151389911677, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.770989418029785, "avg_step_time_sec": 28.437462857791356, "gpu_allocated_gb": 32.365654016, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 398.21754717826843, "eta_sec": 5289.368091549192, "timestamp": "2026-02-15T14:36:29.620330"} |
| {"step": 15, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.64, "mean_correctness_reward": 0.64, "mean_length_penalty": -0.804296875, "mean_total_reward": 0.64, "std_total_reward": 0.48, "total_tokens": 82360, "mean_completion_length": 1647.2, "median_completion_length": 1677, "min_completion_length": 819, "max_completion_length": 2048, "frac_has_think": 0.68, "frac_has_boxed": 0.68, "mean_reward_std_per_prompt": 0.1916515138991168, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.31781029701233, "avg_step_time_sec": 28.429486020406088, "gpu_allocated_gb": 32.36562944, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 426.54192209243774, "eta_sec": 5259.454913775126, "timestamp": "2026-02-15T14:36:58.014568"} |
| {"step": 16, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.52, "mean_correctness_reward": 0.52, "mean_length_penalty": -0.84736328125, "mean_total_reward": 0.52, "std_total_reward": 0.49959983987187184, "total_tokens": 86770, "mean_completion_length": 1735.4, "median_completion_length": 2015, "min_completion_length": 827, "max_completion_length": 2048, "frac_has_think": 0.64, "frac_has_boxed": 0.64, "mean_reward_std_per_prompt": 0.2579795897113271, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 27.80190110206604, "avg_step_time_sec": 28.390261963009834, "gpu_allocated_gb": 32.365674496, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 454.3498454093933, "eta_sec": 5223.8082011938095, "timestamp": "2026-02-15T14:37:26.245532"} |
| {"step": 17, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.58, "mean_correctness_reward": 0.58, "mean_length_penalty": -0.834482421875, "mean_total_reward": 0.58, "std_total_reward": 0.4935585071701226, "total_tokens": 85451, "mean_completion_length": 1709.02, "median_completion_length": 1804, "min_completion_length": 1014, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.82, "mean_reward_std_per_prompt": 0.2433030277982336, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.22400689125061, "avg_step_time_sec": 28.38048225290635, "gpu_allocated_gb": 32.365631488, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 482.580575466156, "eta_sec": 5193.628252281862, "timestamp": "2026-02-15T14:37:54.861905"} |
| {"step": 18, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.46, "mean_correctness_reward": 0.46, "mean_length_penalty": -0.8698828125, "mean_total_reward": 0.46, "std_total_reward": 0.4983974317750845, "total_tokens": 89076, "mean_completion_length": 1781.52, "median_completion_length": 2048, "min_completion_length": 700, "max_completion_length": 2048, "frac_has_think": 0.52, "frac_has_boxed": 0.52, "mean_reward_std_per_prompt": 0.2696311036104439, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.61623740196228, "avg_step_time_sec": 28.393579761187237, "gpu_allocated_gb": 32.36563456, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 511.20310282707214, "eta_sec": 5167.6315165360775, "timestamp": "2026-02-15T14:38:23.403245"} |
| {"step": 19, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.5, "mean_correctness_reward": 0.5, "mean_length_penalty": -0.882021484375, "mean_total_reward": 0.5, "std_total_reward": 0.5, "total_tokens": 90319, "mean_completion_length": 1806.38, "median_completion_length": 2048, "min_completion_length": 744, "max_completion_length": 2048, "frac_has_think": 0.6, "frac_has_boxed": 0.6, "mean_reward_std_per_prompt": 0.36761069332177104, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.591432332992554, "avg_step_time_sec": 28.403993054440146, "gpu_allocated_gb": 32.365650944, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 539.8011744022369, "eta_sec": 5141.122742853667, "timestamp": "2026-02-15T14:38:52.607797"} |
| {"step": 20, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.66, "mean_correctness_reward": 0.66, "mean_length_penalty": -0.8453515625, "mean_total_reward": 0.66, "std_total_reward": 0.4737087712930804, "total_tokens": 86564, "mean_completion_length": 1731.28, "median_completion_length": 1761, "min_completion_length": 1052, "max_completion_length": 2048, "frac_has_think": 0.72, "frac_has_boxed": 0.72, "mean_reward_std_per_prompt": 0.3, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 29.023996829986572, "avg_step_time_sec": 28.43499324321747, "gpu_allocated_gb": 32.365620224, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 568.8316895961761, "eta_sec": 5118.298783779144, "timestamp": "2026-02-15T14:39:20.744027"} |
| {"step": 21, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.68, "mean_correctness_reward": 0.68, "mean_length_penalty": -0.755517578125, "mean_total_reward": 0.68, "std_total_reward": 0.466476151587624, "total_tokens": 77365, "mean_completion_length": 1547.3, "median_completion_length": 1629, "min_completion_length": 698, "max_completion_length": 2048, "frac_has_think": 0.72, "frac_has_boxed": 0.72, "mean_reward_std_per_prompt": 0.2979795897113271, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 28.292150259017944, "avg_step_time_sec": 28.428191196350824, "gpu_allocated_gb": 32.365652992, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 597.1299018859863, "eta_sec": 5088.646224146797, "timestamp": "2026-02-15T14:39:49.108186"} |
| {"step": 22, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.46, "mean_correctness_reward": 0.46, "mean_length_penalty": -0.8401171875, "mean_total_reward": 0.46, "std_total_reward": 0.4983974317750845, "total_tokens": 86028, "mean_completion_length": 1720.56, "median_completion_length": 1883, "min_completion_length": 1002, "max_completion_length": 2048, "frac_has_think": 0.64, "frac_has_boxed": 0.64, "mean_reward_std_per_prompt": 0.3179795897113271, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.895788192749023, "avg_step_time_sec": 28.449445605278015, "gpu_allocated_gb": 32.365769728, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.392025088, "elapsed_sec": 626.0322341918945, "eta_sec": 5064.001317739487, "timestamp": "2026-02-15T14:40:18.242001"} |
| {"step": 23, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.839130859375, "mean_total_reward": 0.4, "std_total_reward": 0.4898979485566356, "total_tokens": 85927, "mean_completion_length": 1718.54, "median_completion_length": 2048, "min_completion_length": 826, "max_completion_length": 2048, "frac_has_think": 0.58, "frac_has_boxed": 0.58, "mean_reward_std_per_prompt": 0.2579795897113271, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 29.472336292266846, "avg_step_time_sec": 28.493919113407966, "gpu_allocated_gb": 32.365832192, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 655.511421918869, "eta_sec": 5043.42368307321, "timestamp": "2026-02-15T14:40:47.946349"} |
| {"step": 24, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.66748046875, "mean_total_reward": 0.78, "std_total_reward": 0.4142463035441596, "total_tokens": 68350, "mean_completion_length": 1367, "median_completion_length": 1180, "min_completion_length": 589, "max_completion_length": 2048, "frac_has_think": 0.78, "frac_has_boxed": 0.78, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.692120790481567, "avg_step_time_sec": 28.502177516619366, "gpu_allocated_gb": 32.365621248, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 684.2103641033173, "eta_sec": 5016.383242925009, "timestamp": "2026-02-15T14:41:14.588522"} |
| {"step": 25, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.52, "mean_correctness_reward": 0.52, "mean_length_penalty": -0.667294921875, "mean_total_reward": 0.52, "std_total_reward": 0.49959983987187184, "total_tokens": 68331, "mean_completion_length": 1366.62, "median_completion_length": 1380, "min_completion_length": 718, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.1516515138991168, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 26.678990840911865, "avg_step_time_sec": 28.429250049591065, "gpu_allocated_gb": 32.365627392, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 710.8961765766144, "eta_sec": 4975.118758678436, "timestamp": "2026-02-15T14:41:41.762037"} |
| {"step": 26, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.750146484375, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 76815, "mean_completion_length": 1536.3, "median_completion_length": 1510, "min_completion_length": 765, "max_completion_length": 2048, "frac_has_think": 0.78, "frac_has_boxed": 0.78, "mean_reward_std_per_prompt": 0.12, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 27.263764142990112, "avg_step_time_sec": 28.38442366856795, "gpu_allocated_gb": 32.365635584, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 738.1659905910492, "eta_sec": 4938.889718330824, "timestamp": "2026-02-15T14:42:09.187669"} |
| {"step": 27, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.76, "mean_correctness_reward": 0.76, "mean_length_penalty": -0.625234375, "mean_total_reward": 0.76, "std_total_reward": 0.4270831300812524, "total_tokens": 64024, "mean_completion_length": 1280.48, "median_completion_length": 1184, "min_completion_length": 661, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.12, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 27.22283148765564, "avg_step_time_sec": 28.34140173594157, "gpu_allocated_gb": 32.3656192, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 765.3950321674347, "eta_sec": 4903.062500317892, "timestamp": "2026-02-15T14:42:35.971312"} |
| {"step": 28, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.58, "mean_correctness_reward": 0.58, "mean_length_penalty": -0.822978515625, "mean_total_reward": 0.58, "std_total_reward": 0.4935585071701226, "total_tokens": 84273, "mean_completion_length": 1685.46, "median_completion_length": 1975, "min_completion_length": 569, "max_completion_length": 2048, "frac_has_think": 0.68, "frac_has_boxed": 0.68, "mean_reward_std_per_prompt": 0.3549545416973504, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 26.85752820968628, "avg_step_time_sec": 28.288406252861023, "gpu_allocated_gb": 32.365636608, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 792.2586114406586, "eta_sec": 4865.605875492096, "timestamp": "2026-02-15T14:43:04.168511"} |
| {"step": 29, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.88, "mean_correctness_reward": 0.88, "mean_length_penalty": -0.655947265625, "mean_total_reward": 0.88, "std_total_reward": 0.32496153618543844, "total_tokens": 67169, "mean_completion_length": 1343.38, "median_completion_length": 1220, "min_completion_length": 746, "max_completion_length": 2048, "frac_has_think": 0.92, "frac_has_boxed": 0.92, "mean_reward_std_per_prompt": 0.16, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 28.510873556137085, "avg_step_time_sec": 28.29607753918089, "gpu_allocated_gb": 32.365682688, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 820.7757291793823, "eta_sec": 4838.629259199932, "timestamp": "2026-02-15T14:43:30.984257"} |
| {"step": 30, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.28, "mean_correctness_reward": 0.28, "mean_length_penalty": -0.80962890625, "mean_total_reward": 0.28, "std_total_reward": 0.44899888641287294, "total_tokens": 82906, "mean_completion_length": 1658.12, "median_completion_length": 1753, "min_completion_length": 642, "max_completion_length": 2048, "frac_has_think": 0.74, "frac_has_boxed": 0.74, "mean_reward_std_per_prompt": 0.2979795897113271, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 26.39392066001892, "avg_step_time_sec": 28.23267230987549, "gpu_allocated_gb": 32.365612032, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 847.1757757663727, "eta_sec": 4799.554292678833, "timestamp": "2026-02-15T14:43:59.374267"} |
| {"step": 31, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.84, "mean_correctness_reward": 0.84, "mean_length_penalty": -0.634716796875, "mean_total_reward": 0.84, "std_total_reward": 0.3666060555964672, "total_tokens": 64995, "mean_completion_length": 1299.9, "median_completion_length": 1196, "min_completion_length": 575, "max_completion_length": 2048, "frac_has_think": 0.88, "frac_has_boxed": 0.88, "mean_reward_std_per_prompt": 0.24, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 28.61302638053894, "avg_step_time_sec": 28.244941796025923, "gpu_allocated_gb": 32.365669376, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 875.7949011325836, "eta_sec": 4773.395163528381, "timestamp": "2026-02-15T14:44:25.826878"} |
| {"step": 32, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.54, "mean_correctness_reward": 0.54, "mean_length_penalty": -0.722392578125, "mean_total_reward": 0.54, "std_total_reward": 0.4983974317750845, "total_tokens": 73973, "mean_completion_length": 1479.46, "median_completion_length": 1431, "min_completion_length": 846, "max_completion_length": 2048, "frac_has_think": 0.76, "frac_has_boxed": 0.76, "mean_reward_std_per_prompt": 0.18, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 26.218950033187866, "avg_step_time_sec": 28.181629553437233, "gpu_allocated_gb": 32.365612032, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 902.0204594135284, "eta_sec": 4734.513764977455, "timestamp": "2026-02-15T14:44:53.486350"} |
| {"step": 33, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.7, "mean_correctness_reward": 0.7, "mean_length_penalty": -0.73466796875, "mean_total_reward": 0.7, "std_total_reward": 0.458257569495584, "total_tokens": 75230, "mean_completion_length": 1504.6, "median_completion_length": 1419, "min_completion_length": 760, "max_completion_length": 2048, "frac_has_think": 0.76, "frac_has_boxed": 0.76, "mean_reward_std_per_prompt": 0.26, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 27.88970923423767, "avg_step_time_sec": 28.172783483158458, "gpu_allocated_gb": 32.365663232, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 929.9162635803223, "eta_sec": 4704.854841687463, "timestamp": "2026-02-15T14:45:20.925570"} |
| {"step": 34, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.688798828125, "mean_total_reward": 0.78, "std_total_reward": 0.4142463035441596, "total_tokens": 70533, "mean_completion_length": 1410.66, "median_completion_length": 1317, "min_completion_length": 805, "max_completion_length": 2048, "frac_has_think": 0.82, "frac_has_boxed": 0.82, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 27.641254425048828, "avg_step_time_sec": 28.157150275567, "gpu_allocated_gb": 32.365696, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 957.5641129016876, "eta_sec": 4674.086945744121, "timestamp": "2026-02-15T14:45:48.631632"} |
| {"step": 35, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.6365625, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 65184, "mean_completion_length": 1303.68, "median_completion_length": 1170, "min_completion_length": 643, "max_completion_length": 2048, "frac_has_think": 0.9, "frac_has_boxed": 0.9, "mean_reward_std_per_prompt": 0.12, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 27.377490997314453, "avg_step_time_sec": 28.134874296188354, "gpu_allocated_gb": 32.365632512, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 984.9480686187744, "eta_sec": 4642.2542588710785, "timestamp": "2026-02-15T14:46:15.097722"} |
| {"step": 36, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.54, "mean_correctness_reward": 0.54, "mean_length_penalty": -0.7208203125, "mean_total_reward": 0.54, "std_total_reward": 0.4983974317750845, "total_tokens": 73812, "mean_completion_length": 1476.24, "median_completion_length": 1418, "min_completion_length": 809, "max_completion_length": 2048, "frac_has_think": 0.76, "frac_has_boxed": 0.76, "mean_reward_std_per_prompt": 0.14, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 26.53976535797119, "avg_step_time_sec": 28.090565714571213, "gpu_allocated_gb": 32.365651968, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1011.4937646389008, "eta_sec": 4606.852777189679, "timestamp": "2026-02-15T14:46:41.666602"} |
| {"step": 37, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.82, "mean_correctness_reward": 0.82, "mean_length_penalty": -0.72123046875, "mean_total_reward": 0.82, "std_total_reward": 0.3841874542459709, "total_tokens": 73854, "mean_completion_length": 1477.08, "median_completion_length": 1657, "min_completion_length": 735, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.2116515138991168, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 27.50022554397583, "avg_step_time_sec": 28.07461057482539, "gpu_allocated_gb": 32.36573184, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1039.000717163086, "eta_sec": 4576.161523696538, "timestamp": "2026-02-15T14:47:09.039355"} |
| {"step": 38, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.56, "mean_correctness_reward": 0.56, "mean_length_penalty": -0.608046875, "mean_total_reward": 0.56, "std_total_reward": 0.4963869458396343, "total_tokens": 62264, "mean_completion_length": 1245.28, "median_completion_length": 1069, "min_completion_length": 576, "max_completion_length": 2048, "frac_has_think": 0.78, "frac_has_boxed": 0.78, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 26.437113523483276, "avg_step_time_sec": 28.03151854715849, "gpu_allocated_gb": 32.365658112, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1065.4465701580048, "eta_sec": 4541.106004639675, "timestamp": "2026-02-15T14:47:34.496371"} |
| {"step": 39, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.48669921875, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 49838, "mean_completion_length": 996.76, "median_completion_length": 935, "min_completion_length": 679, "max_completion_length": 1863, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.222470998764038, "avg_step_time_sec": 27.95949168694325, "gpu_allocated_gb": 32.365620224, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1090.6771273612976, "eta_sec": 4501.478161597863, "timestamp": "2026-02-15T14:47:57.911779"} |
| {"step": 40, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.44, "mean_correctness_reward": 0.44, "mean_length_penalty": -0.662021484375, "mean_total_reward": 0.44, "std_total_reward": 0.4963869458396343, "total_tokens": 67791, "mean_completion_length": 1355.82, "median_completion_length": 1249, "min_completion_length": 602, "max_completion_length": 2048, "frac_has_think": 0.86, "frac_has_boxed": 0.86, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 22.77688241004944, "avg_step_time_sec": 27.829926455020903, "gpu_allocated_gb": 32.365362176, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1113.4603626728058, "eta_sec": 4452.788232803345, "timestamp": "2026-02-15T14:48:23.010372"} |
| {"step": 41, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.66318359375, "mean_total_reward": 0.74, "std_total_reward": 0.4386342439892262, "total_tokens": 67910, "mean_completion_length": 1358.2, "median_completion_length": 1306, "min_completion_length": 584, "max_completion_length": 2048, "frac_has_think": 0.78, "frac_has_boxed": 0.78, "mean_reward_std_per_prompt": 0.1579795897113271, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 26.217646837234497, "avg_step_time_sec": 27.790602561904162, "gpu_allocated_gb": 32.365696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1139.684489250183, "eta_sec": 4418.705807342762, "timestamp": "2026-02-15T14:48:49.229522"} |
| {"step": 42, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.51048828125, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 52274, "mean_completion_length": 1045.48, "median_completion_length": 991, "min_completion_length": 568, "max_completion_length": 1853, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.90692949295044, "avg_step_time_sec": 27.745753203119552, "gpu_allocated_gb": 32.365651968, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1165.5981812477112, "eta_sec": 4383.829006092889, "timestamp": "2026-02-15T14:49:13.000516"} |
| {"step": 43, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.9, "mean_correctness_reward": 0.9, "mean_length_penalty": -0.63216796875, "mean_total_reward": 0.9, "std_total_reward": 0.3, "total_tokens": 64734, "mean_completion_length": 1294.68, "median_completion_length": 1145, "min_completion_length": 739, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.17165151389911681, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 22.691683769226074, "avg_step_time_sec": 27.62821670465691, "gpu_allocated_gb": 32.365333504, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1188.2957026958466, "eta_sec": 4337.630022631135, "timestamp": "2026-02-15T14:49:37.518095"} |
| {"step": 44, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.72, "mean_correctness_reward": 0.72, "mean_length_penalty": -0.575732421875, "mean_total_reward": 0.72, "std_total_reward": 0.44899888641287294, "total_tokens": 58955, "mean_completion_length": 1179.1, "median_completion_length": 1072, "min_completion_length": 482, "max_completion_length": 2048, "frac_has_think": 0.92, "frac_has_boxed": 0.92, "mean_reward_std_per_prompt": 0.09797958971132711, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.547446250915527, "avg_step_time_sec": 27.58092646707188, "gpu_allocated_gb": 32.365645824, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1213.8498644828796, "eta_sec": 4302.6245288632135, "timestamp": "2026-02-15T14:50:02.624692"} |
| {"step": 45, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.620322265625, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 63521, "mean_completion_length": 1270.42, "median_completion_length": 1112, "min_completion_length": 684, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.585453271865845, "avg_step_time_sec": 27.536582618289522, "gpu_allocated_gb": 32.365638656, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1239.4420971870422, "eta_sec": 4268.170305834876, "timestamp": "2026-02-15T14:50:28.604009"} |
| {"step": 46, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.88, "mean_correctness_reward": 0.88, "mean_length_penalty": -0.507734375, "mean_total_reward": 0.88, "std_total_reward": 0.32496153618543844, "total_tokens": 51992, "mean_completion_length": 1039.84, "median_completion_length": 830, "min_completion_length": 615, "max_completion_length": 2048, "frac_has_think": 0.92, "frac_has_boxed": 0.92, "mean_reward_std_per_prompt": 0.16, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.540881872177124, "avg_step_time_sec": 27.493197819460992, "gpu_allocated_gb": 32.3656704, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1264.9891238212585, "eta_sec": 4233.952464196993, "timestamp": "2026-02-15T14:50:53.151129"} |
| {"step": 47, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.66, "mean_correctness_reward": 0.66, "mean_length_penalty": -0.596181640625, "mean_total_reward": 0.66, "std_total_reward": 0.4737087712930804, "total_tokens": 61049, "mean_completion_length": 1220.98, "median_completion_length": 1080, "min_completion_length": 516, "max_completion_length": 2048, "frac_has_think": 0.9, "frac_has_boxed": 0.9, "mean_reward_std_per_prompt": 0.18, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.477508783340454, "avg_step_time_sec": 27.42903422294779, "gpu_allocated_gb": 32.36563456, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1289.473263502121, "eta_sec": 4196.642236111012, "timestamp": "2026-02-15T14:51:18.981310"} |
| {"step": 48, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.576337890625, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 59017, "mean_completion_length": 1180.34, "median_completion_length": 1061, "min_completion_length": 558, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.82827639579773, "avg_step_time_sec": 27.395685101548832, "gpu_allocated_gb": 32.3656448, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1315.3078989982605, "eta_sec": 4164.144135435423, "timestamp": "2026-02-15T14:51:44.037309"} |
| {"step": 49, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.543037109375, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 55607, "mean_completion_length": 1112.14, "median_completion_length": 945, "min_completion_length": 435, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 25.02954411506653, "avg_step_time_sec": 27.347396509987966, "gpu_allocated_gb": 32.365645824, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1340.3439452648163, "eta_sec": 4129.456873008183, "timestamp": "2026-02-15T14:52:08.834647"} |
| {"step": 50, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.51400390625, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 52634, "mean_completion_length": 1052.68, "median_completion_length": 925, "min_completion_length": 606, "max_completion_length": 1869, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.792707443237305, "avg_step_time_sec": 27.296302728652954, "gpu_allocated_gb": 32.365627392, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1365.1431646347046, "eta_sec": 4094.445409297943, "timestamp": "2026-02-15T14:52:33.353122"} |
| {"step": 51, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.483388671875, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 49499, "mean_completion_length": 989.98, "median_completion_length": 900, "min_completion_length": 618, "max_completion_length": 1939, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 23.170626401901245, "avg_step_time_sec": 27.21540711440292, "gpu_allocated_gb": 32.365372928, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1388.8643074035645, "eta_sec": 4055.0956600460354, "timestamp": "2026-02-15T14:52:55.971460"} |
| {"step": 52, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.64, "mean_correctness_reward": 0.64, "mean_length_penalty": -0.615830078125, "mean_total_reward": 0.64, "std_total_reward": 0.48, "total_tokens": 63061, "mean_completion_length": 1261.22, "median_completion_length": 1241, "min_completion_length": 567, "max_completion_length": 2048, "frac_has_think": 0.88, "frac_has_boxed": 0.88, "mean_reward_std_per_prompt": 0.17797958971132713, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 22.81310796737671, "avg_step_time_sec": 27.130747515421646, "gpu_allocated_gb": 32.365469696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1411.6842987537384, "eta_sec": 4015.3506322824037, "timestamp": "2026-02-15T14:53:20.787648"} |
| {"step": 53, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.72, "mean_correctness_reward": 0.72, "mean_length_penalty": -0.65998046875, "mean_total_reward": 0.72, "std_total_reward": 0.44899888641287294, "total_tokens": 67582, "mean_completion_length": 1351.64, "median_completion_length": 1348, "min_completion_length": 669, "max_completion_length": 2048, "frac_has_think": 0.82, "frac_has_boxed": 0.82, "mean_reward_std_per_prompt": 0.3116515138991168, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.410612106323242, "avg_step_time_sec": 27.09829213034432, "gpu_allocated_gb": 32.365645824, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1437.1017162799835, "eta_sec": 3983.448943160615, "timestamp": "2026-02-15T14:53:47.200768"} |
| {"step": 54, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.68, "mean_correctness_reward": 0.68, "mean_length_penalty": -0.67669921875, "mean_total_reward": 0.68, "std_total_reward": 0.466476151587624, "total_tokens": 69294, "mean_completion_length": 1385.88, "median_completion_length": 1280, "min_completion_length": 627, "max_completion_length": 2048, "frac_has_think": 0.7, "frac_has_boxed": 0.7, "mean_reward_std_per_prompt": 0.09797958971132711, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 26.47559928894043, "avg_step_time_sec": 27.086760781429433, "gpu_allocated_gb": 32.365657088, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1463.5840110778809, "eta_sec": 3954.667074088697, "timestamp": "2026-02-15T14:54:13.241155"} |
| {"step": 55, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.548623046875, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 56179, "mean_completion_length": 1123.58, "median_completion_length": 830, "min_completion_length": 463, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 25.97816777229309, "avg_step_time_sec": 27.06660454489968, "gpu_allocated_gb": 32.365641728, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1489.5686905384064, "eta_sec": 3924.6576590104532, "timestamp": "2026-02-15T14:54:38.173733"} |
| {"step": 56, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.7, "mean_correctness_reward": 0.7, "mean_length_penalty": -0.64818359375, "mean_total_reward": 0.7, "std_total_reward": 0.458257569495584, "total_tokens": 66374, "mean_completion_length": 1327.48, "median_completion_length": 984, "min_completion_length": 586, "max_completion_length": 2048, "frac_has_think": 0.78, "frac_has_boxed": 0.78, "mean_reward_std_per_prompt": 0.1, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.411571502685547, "avg_step_time_sec": 27.03705038343157, "gpu_allocated_gb": 32.365648896, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1514.9869525432587, "eta_sec": 3893.335255214146, "timestamp": "2026-02-15T14:55:04.394290"} |
| {"step": 57, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.74623046875, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 76414, "mean_completion_length": 1528.28, "median_completion_length": 1735, "min_completion_length": 615, "max_completion_length": 2048, "frac_has_think": 0.62, "frac_has_boxed": 0.62, "mean_reward_std_per_prompt": 0.12, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.716461896896362, "avg_step_time_sec": 27.013882164369548, "gpu_allocated_gb": 32.365647872, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1540.710158109665, "eta_sec": 3862.985149504845, "timestamp": "2026-02-15T14:55:31.163001"} |
| {"step": 58, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.84, "mean_correctness_reward": 0.84, "mean_length_penalty": -0.56099609375, "mean_total_reward": 0.84, "std_total_reward": 0.3666060555964672, "total_tokens": 57446, "mean_completion_length": 1148.92, "median_completion_length": 1119, "min_completion_length": 523, "max_completion_length": 2048, "frac_has_think": 0.92, "frac_has_boxed": 0.92, "mean_reward_std_per_prompt": 0.1516515138991168, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 27.23799204826355, "avg_step_time_sec": 27.017746127884962, "gpu_allocated_gb": 32.36573696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1567.955235004425, "eta_sec": 3836.5199501596644, "timestamp": "2026-02-15T14:55:57.082257"} |
| {"step": 59, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.53259765625, "mean_total_reward": 0.4, "std_total_reward": 0.4898979485566356, "total_tokens": 54538, "mean_completion_length": 1090.76, "median_completion_length": 1182, "min_completion_length": 573, "max_completion_length": 1517, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 25.38019895553589, "avg_step_time_sec": 26.989991091065487, "gpu_allocated_gb": 32.365630464, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1593.341781616211, "eta_sec": 3805.5887438402337, "timestamp": "2026-02-15T14:56:18.446590"} |
| {"step": 60, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.475341796875, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 48675, "mean_completion_length": 973.5, "median_completion_length": 926, "min_completion_length": 561, "max_completion_length": 1565, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 19.32543706893921, "avg_step_time_sec": 26.86224852403005, "gpu_allocated_gb": 32.364909056, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1612.6733345985413, "eta_sec": 3760.7147933642073, "timestamp": "2026-02-15T14:56:37.466437"} |
| {"step": 61, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.76, "mean_correctness_reward": 0.76, "mean_length_penalty": -0.672392578125, "mean_total_reward": 0.76, "std_total_reward": 0.4270831300812524, "total_tokens": 68853, "mean_completion_length": 1377.06, "median_completion_length": 1406, "min_completion_length": 518, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.2633030277982336, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 18.963560342788696, "avg_step_time_sec": 26.73276183253429, "gpu_allocated_gb": 32.364945408, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1631.6437118053436, "eta_sec": 3715.8538947222664, "timestamp": "2026-02-15T14:57:01.817886"} |
| {"step": 62, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.82, "mean_correctness_reward": 0.82, "mean_length_penalty": -0.489208984375, "mean_total_reward": 0.82, "std_total_reward": 0.3841874542459709, "total_tokens": 50095, "mean_completion_length": 1001.9, "median_completion_length": 925, "min_completion_length": 437, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 26.55996608734131, "avg_step_time_sec": 26.729974804386014, "gpu_allocated_gb": 32.365659136, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1658.209977388382, "eta_sec": 3688.73652300527, "timestamp": "2026-02-15T14:57:26.237581"} |
| {"step": 63, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.5284765625, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 54116, "mean_completion_length": 1082.32, "median_completion_length": 1074, "min_completion_length": 648, "max_completion_length": 1856, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.4080069065094, "avg_step_time_sec": 26.693118171086386, "gpu_allocated_gb": 32.365652992, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1682.6243977546692, "eta_sec": 3656.957189438835, "timestamp": "2026-02-15T14:57:49.661364"} |
| {"step": 64, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.86, "mean_correctness_reward": 0.86, "mean_length_penalty": -0.56791015625, "mean_total_reward": 0.86, "std_total_reward": 0.34698703145794946, "total_tokens": 58154, "mean_completion_length": 1163.08, "median_completion_length": 1083, "min_completion_length": 545, "max_completion_length": 2048, "frac_has_think": 0.92, "frac_has_boxed": 0.92, "mean_reward_std_per_prompt": 0.1579795897113271, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 22.55530858039856, "avg_step_time_sec": 26.62846489623189, "gpu_allocated_gb": 32.365374976, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1705.186535835266, "eta_sec": 3621.471225887537, "timestamp": "2026-02-15T14:58:14.438927"} |
| {"step": 65, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.76, "mean_correctness_reward": 0.76, "mean_length_penalty": -0.5147265625, "mean_total_reward": 0.76, "std_total_reward": 0.4270831300812524, "total_tokens": 52708, "mean_completion_length": 1054.16, "median_completion_length": 874, "min_completion_length": 426, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.63929510116577, "avg_step_time_sec": 26.61324689938472, "gpu_allocated_gb": 32.36566016, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1730.832322359085, "eta_sec": 3592.788331416937, "timestamp": "2026-02-15T14:58:39.184296"} |
| {"step": 66, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.58, "mean_correctness_reward": 0.58, "mean_length_penalty": -0.517255859375, "mean_total_reward": 0.58, "std_total_reward": 0.4935585071701226, "total_tokens": 52967, "mean_completion_length": 1059.34, "median_completion_length": 1035, "min_completion_length": 351, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 24.716274976730347, "avg_step_time_sec": 26.584504900556624, "gpu_allocated_gb": 32.36565504, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1755.5553011894226, "eta_sec": 3562.3236566745877, "timestamp": "2026-02-15T14:59:03.609500"} |
| {"step": 67, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.549814453125, "mean_total_reward": 0.78, "std_total_reward": 0.4142463035441596, "total_tokens": 56301, "mean_completion_length": 1126.02, "median_completion_length": 999, "min_completion_length": 688, "max_completion_length": 2048, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.823358058929443, "avg_step_time_sec": 26.558219126800992, "gpu_allocated_gb": 32.365636608, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1780.3861801624298, "eta_sec": 3532.243143864532, "timestamp": "2026-02-15T14:59:28.877974"} |
| {"step": 68, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.542705078125, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 55573, "mean_completion_length": 1111.46, "median_completion_length": 876, "min_completion_length": 484, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 24.819190502166748, "avg_step_time_sec": 26.532645176438724, "gpu_allocated_gb": 32.365647872, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1805.2115502357483, "eta_sec": 3502.3091632899113, "timestamp": "2026-02-15T14:59:53.632409"} |
| {"step": 69, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.22, "mean_correctness_reward": 0.22, "mean_length_penalty": -0.63111328125, "mean_total_reward": 0.22, "std_total_reward": 0.4142463035441596, "total_tokens": 64626, "mean_completion_length": 1292.52, "median_completion_length": 987, "min_completion_length": 572, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 24.760334968566895, "avg_step_time_sec": 26.506959521252178, "gpu_allocated_gb": 32.365641728, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1829.9779090881348, "eta_sec": 3472.411697284035, "timestamp": "2026-02-15T15:00:19.314539"} |
| {"step": 70, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.58, "mean_correctness_reward": 0.58, "mean_length_penalty": -0.474111328125, "mean_total_reward": 0.58, "std_total_reward": 0.4935585071701226, "total_tokens": 48549, "mean_completion_length": 970.98, "median_completion_length": 862, "min_completion_length": 521, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 26.17525029182434, "avg_step_time_sec": 26.502220817974635, "gpu_allocated_gb": 32.36564992, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1856.1601123809814, "eta_sec": 3445.2887063367025, "timestamp": "2026-02-15T15:00:43.959743"} |
| {"step": 71, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.4387890625, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 44932, "mean_completion_length": 898.64, "median_completion_length": 813, "min_completion_length": 632, "max_completion_length": 1741, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 23.97816514968872, "avg_step_time_sec": 26.466670738139623, "gpu_allocated_gb": 32.365621248, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1880.1446554660797, "eta_sec": 3414.2005252200115, "timestamp": "2026-02-15T15:01:05.725817"} |
| {"step": 72, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.566640625, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 58024, "mean_completion_length": 1160.48, "median_completion_length": 975, "min_completion_length": 620, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 20.677480459213257, "avg_step_time_sec": 26.38626531759898, "gpu_allocated_gb": 32.365236736, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1900.828957080841, "eta_sec": 3377.4419606526694, "timestamp": "2026-02-15T15:01:29.429944"} |
| {"step": 73, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.523798828125, "mean_total_reward": 0.4, "std_total_reward": 0.4898979485566356, "total_tokens": 53637, "mean_completion_length": 1072.74, "median_completion_length": 937, "min_completion_length": 541, "max_completion_length": 2048, "frac_has_think": 0.86, "frac_has_boxed": 0.86, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 25.44390368461609, "avg_step_time_sec": 26.37335625413346, "gpu_allocated_gb": 32.36562944, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1926.2794439792633, "eta_sec": 3349.4162442749493, "timestamp": "2026-02-15T15:01:54.637983"} |
| {"step": 74, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.86, "mean_correctness_reward": 0.86, "mean_length_penalty": -0.54169921875, "mean_total_reward": 0.86, "std_total_reward": 0.34698703145794946, "total_tokens": 55470, "mean_completion_length": 1109.4, "median_completion_length": 924, "min_completion_length": 503, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.0916515138991168, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.70974588394165, "avg_step_time_sec": 26.35087503291465, "gpu_allocated_gb": 32.365658112, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1950.995854139328, "eta_sec": 3320.210254147246, "timestamp": "2026-02-15T15:02:19.472268"} |
| {"step": 75, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.509736328125, "mean_total_reward": 0.78, "std_total_reward": 0.4142463035441596, "total_tokens": 52197, "mean_completion_length": 1043.94, "median_completion_length": 953, "min_completion_length": 258, "max_completion_length": 2048, "frac_has_think": 0.84, "frac_has_boxed": 0.84, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.76156497001648, "avg_step_time_sec": 26.32968423207601, "gpu_allocated_gb": 32.365628416, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1975.7636942863464, "eta_sec": 3291.210529009501, "timestamp": "2026-02-15T15:02:44.019183"} |
| {"step": 76, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.84, "mean_correctness_reward": 0.84, "mean_length_penalty": -0.51294921875, "mean_total_reward": 0.84, "std_total_reward": 0.3666060555964672, "total_tokens": 52526, "mean_completion_length": 1050.52, "median_completion_length": 798, "min_completion_length": 461, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.622846126556396, "avg_step_time_sec": 26.30722583595075, "gpu_allocated_gb": 32.365657088, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2000.3935272693634, "eta_sec": 3262.096003657893, "timestamp": "2026-02-15T15:03:09.226762"} |
| {"step": 77, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.51205078125, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 52434, "mean_completion_length": 1048.68, "median_completion_length": 934, "min_completion_length": 534, "max_completion_length": 2048, "frac_has_think": 0.92, "frac_has_boxed": 0.92, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 25.109015703201294, "avg_step_time_sec": 26.29166466539556, "gpu_allocated_gb": 32.36562432, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2025.5095000267029, "eta_sec": 3233.874753843654, "timestamp": "2026-02-15T15:03:33.665567"} |
| {"step": 78, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.54353515625, "mean_total_reward": 0.78, "std_total_reward": 0.4142463035441596, "total_tokens": 55658, "mean_completion_length": 1113.16, "median_completion_length": 1085, "min_completion_length": 603, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.454113483428955, "avg_step_time_sec": 26.268106316908813, "gpu_allocated_gb": 32.365625344, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2049.970621109009, "eta_sec": 3204.7089706628753, "timestamp": "2026-02-15T15:03:58.249555"} |
| {"step": 79, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.489482421875, "mean_total_reward": 0.4, "std_total_reward": 0.4898979485566356, "total_tokens": 50123, "mean_completion_length": 1002.46, "median_completion_length": 968, "min_completion_length": 649, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 25.218411445617676, "avg_step_time_sec": 26.254819040057026, "gpu_allocated_gb": 32.365647872, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2075.19544506073, "eta_sec": 3176.8331038469, "timestamp": "2026-02-15T15:04:23.070186"} |
| {"step": 80, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.584638671875, "mean_total_reward": 0.74, "std_total_reward": 0.4386342439892262, "total_tokens": 59867, "mean_completion_length": 1197.34, "median_completion_length": 1210, "min_completion_length": 519, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.0916515138991168, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.25397515296936, "avg_step_time_sec": 26.22980849146843, "gpu_allocated_gb": 32.365650944, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2099.455872297287, "eta_sec": 3147.5770189762115, "timestamp": "2026-02-15T15:04:48.435351"} |
| {"step": 81, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.399287109375, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 40887, "mean_completion_length": 817.74, "median_completion_length": 790, "min_completion_length": 472, "max_completion_length": 1238, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.72161602973938, "avg_step_time_sec": 26.22353451045943, "gpu_allocated_gb": 32.365735936, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2125.1865241527557, "eta_sec": 3120.6006067446724, "timestamp": "2026-02-15T15:05:07.468577"} |
| {"step": 82, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.56, "mean_correctness_reward": 0.56, "mean_length_penalty": -0.45947265625, "mean_total_reward": 0.56, "std_total_reward": 0.4963869458396343, "total_tokens": 47050, "mean_completion_length": 941, "median_completion_length": 836, "min_completion_length": 585, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.12, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 15.270360708236694, "avg_step_time_sec": 26.08995922018842, "gpu_allocated_gb": 32.364504064, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2140.463709831238, "eta_sec": 3078.6151879822332, "timestamp": "2026-02-15T15:05:28.436372"} |
| {"step": 83, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.466611328125, "mean_total_reward": 0.78, "std_total_reward": 0.4142463035441596, "total_tokens": 47781, "mean_completion_length": 955.62, "median_completion_length": 740, "min_completion_length": 411, "max_completion_length": 2048, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.21524429321289, "avg_step_time_sec": 26.06737229335739, "gpu_allocated_gb": 32.365630464, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2164.6851530075073, "eta_sec": 3049.8825583228145, "timestamp": "2026-02-15T15:05:52.532978"} |
| {"step": 84, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.44, "mean_correctness_reward": 0.44, "mean_length_penalty": -0.606181640625, "mean_total_reward": 0.44, "std_total_reward": 0.4963869458396343, "total_tokens": 62073, "mean_completion_length": 1241.46, "median_completion_length": 857, "min_completion_length": 437, "max_completion_length": 2048, "frac_has_think": 0.76, "frac_has_boxed": 0.78, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 24.213375091552734, "avg_step_time_sec": 26.04530089809781, "gpu_allocated_gb": 32.36566016, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2188.905522108078, "eta_sec": 3021.254904179346, "timestamp": "2026-02-15T15:06:17.832404"} |
| {"step": 85, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.33259765625, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 34058, "mean_completion_length": 681.16, "median_completion_length": 677, "min_completion_length": 290, "max_completion_length": 1246, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.073232650756836, "avg_step_time_sec": 26.03386480107027, "gpu_allocated_gb": 32.365607936, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2213.984860420227, "eta_sec": 2993.894452123081, "timestamp": "2026-02-15T15:06:36.238993"} |
| {"step": 86, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.488046875, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 49976, "mean_completion_length": 999.52, "median_completion_length": 1008, "min_completion_length": 313, "max_completion_length": 1871, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 15.180322885513306, "avg_step_time_sec": 25.90766082530798, "gpu_allocated_gb": 32.364491264, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2229.1716117858887, "eta_sec": 2953.4733340851094, "timestamp": "2026-02-15T15:06:55.995534"} |
| {"step": 87, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.381640625, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 39080, "mean_completion_length": 781.6, "median_completion_length": 768, "min_completion_length": 453, "max_completion_length": 1286, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 22.426785469055176, "avg_step_time_sec": 25.867650763741857, "gpu_allocated_gb": 32.365371904, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2251.6045315265656, "eta_sec": 2923.04453630283, "timestamp": "2026-02-15T15:07:13.995886"} |
| {"step": 88, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.52447265625, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 53706, "mean_completion_length": 1074.12, "median_completion_length": 968, "min_completion_length": 447, "max_completion_length": 2048, "frac_has_think": 0.86, "frac_has_boxed": 0.86, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 15.585855484008789, "avg_step_time_sec": 25.750812181017615, "gpu_allocated_gb": 32.364588544, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2267.1971793174744, "eta_sec": 2884.090964273973, "timestamp": "2026-02-15T15:07:36.095743"} |
| {"step": 89, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.62814453125, "mean_total_reward": 0.74, "std_total_reward": 0.4386342439892262, "total_tokens": 64322, "mean_completion_length": 1286.44, "median_completion_length": 1180, "min_completion_length": 569, "max_completion_length": 2048, "frac_has_think": 0.88, "frac_has_boxed": 0.88, "mean_reward_std_per_prompt": 0.1579795897113271, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.27887272834778, "avg_step_time_sec": 25.74550949053818, "gpu_allocated_gb": 32.365671424, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2292.4822561740875, "eta_sec": 2857.751553449738, "timestamp": "2026-02-15T15:08:01.647941"} |
| {"step": 90, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.82, "mean_correctness_reward": 0.82, "mean_length_penalty": -0.537568359375, "mean_total_reward": 0.82, "std_total_reward": 0.3841874542459709, "total_tokens": 55047, "mean_completion_length": 1100.94, "median_completion_length": 844, "min_completion_length": 359, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.480966329574585, "avg_step_time_sec": 25.742570122083027, "gpu_allocated_gb": 32.365637632, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2317.969828605652, "eta_sec": 2831.682713429133, "timestamp": "2026-02-15T15:08:26.677937"} |
| {"step": 91, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.62, "mean_correctness_reward": 0.62, "mean_length_penalty": -0.513701171875, "mean_total_reward": 0.62, "std_total_reward": 0.4853864439804639, "total_tokens": 52603, "mean_completion_length": 1052.06, "median_completion_length": 859, "min_completion_length": 456, "max_completion_length": 2048, "frac_has_think": 0.92, "frac_has_boxed": 0.92, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.78880286216736, "avg_step_time_sec": 25.743078174171867, "gpu_allocated_gb": 32.365791232, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2343.765125989914, "eta_sec": 2805.9955209847335, "timestamp": "2026-02-15T15:08:52.508282"} |
| {"step": 92, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.68, "mean_correctness_reward": 0.68, "mean_length_penalty": -0.6905078125, "mean_total_reward": 0.68, "std_total_reward": 0.466476151587624, "total_tokens": 70708, "mean_completion_length": 1414.16, "median_completion_length": 1416, "min_completion_length": 650, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.09797958971132711, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.067942142486572, "avg_step_time_sec": 25.735739739044853, "gpu_allocated_gb": 32.365626368, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2368.839988708496, "eta_sec": 2779.459891816844, "timestamp": "2026-02-15T15:09:18.810401"} |
| {"step": 93, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.56, "mean_correctness_reward": 0.56, "mean_length_penalty": -0.51544921875, "mean_total_reward": 0.56, "std_total_reward": 0.4963869458396343, "total_tokens": 52782, "mean_completion_length": 1055.64, "median_completion_length": 881, "min_completion_length": 579, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 26.860814809799194, "avg_step_time_sec": 25.747837320450813, "gpu_allocated_gb": 32.365775872, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2395.7076177597046, "eta_sec": 2755.018593288237, "timestamp": "2026-02-15T15:09:44.163227"} |
| {"step": 94, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.96, "mean_correctness_reward": 0.96, "mean_length_penalty": -0.434814453125, "mean_total_reward": 0.96, "std_total_reward": 0.19595917942265426, "total_tokens": 44525, "mean_completion_length": 890.5, "median_completion_length": 825, "min_completion_length": 403, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.638973236083984, "avg_step_time_sec": 25.74667919189372, "gpu_allocated_gb": 32.365821952, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2421.3538591861725, "eta_sec": 2729.1479943407344, "timestamp": "2026-02-15T15:10:09.338985"} |
| {"step": 95, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.49841796875, "mean_total_reward": 0.78, "std_total_reward": 0.4142463035441596, "total_tokens": 51038, "mean_completion_length": 1020.76, "median_completion_length": 956, "min_completion_length": 608, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.16193723678589, "avg_step_time_sec": 25.729997697629425, "gpu_allocated_gb": 32.365621248, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2445.522100687027, "eta_sec": 2701.6497582510897, "timestamp": "2026-02-15T15:10:33.526124"} |
| {"step": 96, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.94, "mean_correctness_reward": 0.94, "mean_length_penalty": -0.46751953125, "mean_total_reward": 0.94, "std_total_reward": 0.23748684174075835, "total_tokens": 47874, "mean_completion_length": 957.48, "median_completion_length": 873, "min_completion_length": 499, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.14, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.391906023025513, "avg_step_time_sec": 25.716059242685635, "gpu_allocated_gb": 32.365652992, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2469.9210107326508, "eta_sec": 2674.470161239306, "timestamp": "2026-02-15T15:10:57.654304"} |
| {"step": 97, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.429609375, "mean_total_reward": 0.4, "std_total_reward": 0.4898979485566356, "total_tokens": 43992, "mean_completion_length": 879.84, "median_completion_length": 666, "min_completion_length": 396, "max_completion_length": 2048, "frac_has_think": 0.84, "frac_has_boxed": 0.84, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 24.04356551170349, "avg_step_time_sec": 25.69881703927345, "gpu_allocated_gb": 32.365643776, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2493.971355199814, "eta_sec": 2646.9781550451653, "timestamp": "2026-02-15T15:11:22.056093"} |
| {"step": 98, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.35533203125, "mean_total_reward": 0.74, "std_total_reward": 0.4386342439892262, "total_tokens": 36386, "mean_completion_length": 727.72, "median_completion_length": 759, "min_completion_length": 457, "max_completion_length": 1066, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0916515138991168, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.378324031829834, "avg_step_time_sec": 25.685342620830145, "gpu_allocated_gb": 32.365626368, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2518.3576605319977, "eta_sec": 2619.9049473246746, "timestamp": "2026-02-15T15:11:39.236007"} |
| {"step": 99, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.34177734375, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 34998, "mean_completion_length": 699.96, "median_completion_length": 674, "min_completion_length": 378, "max_completion_length": 1031, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 13.295036315917969, "avg_step_time_sec": 25.56018801168962, "gpu_allocated_gb": 32.364268032, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2531.659415245056, "eta_sec": 2581.5789891806517, "timestamp": "2026-02-15T15:11:52.194751"} |
| {"step": 100, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.45298828125, "mean_total_reward": 0.98, "std_total_reward": 0.14, "total_tokens": 46386, "mean_completion_length": 927.72, "median_completion_length": 861, "min_completion_length": 570, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 12.673522233963013, "avg_step_time_sec": 25.431321353912352, "gpu_allocated_gb": 32.36419328, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2544.3392567634583, "eta_sec": 2543.1321353912354, "timestamp": "2026-02-15T15:12:12.207379"} |
| {"step": 101, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.96, "mean_correctness_reward": 0.96, "mean_length_penalty": -0.458212890625, "mean_total_reward": 0.96, "std_total_reward": 0.19595917942265426, "total_tokens": 46921, "mean_completion_length": 938.42, "median_completion_length": 764, "min_completion_length": 421, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.07634401321411, "avg_step_time_sec": 25.417905736677717, "gpu_allocated_gb": 32.365668352, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2569.008684396744, "eta_sec": 2516.372667931094, "timestamp": "2026-02-15T15:12:36.571582"} |
| {"step": 102, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.380166015625, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 38929, "mean_completion_length": 778.58, "median_completion_length": 771, "min_completion_length": 426, "max_completion_length": 1378, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 23.93232250213623, "avg_step_time_sec": 25.403341195162604, "gpu_allocated_gb": 32.365659136, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2592.948930978775, "eta_sec": 2489.527437125935, "timestamp": "2026-02-15T15:12:56.054231"} |
| {"step": 103, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.36, "mean_correctness_reward": 0.36, "mean_length_penalty": -0.6646875, "mean_total_reward": 0.36, "std_total_reward": 0.48, "total_tokens": 68064, "mean_completion_length": 1361.28, "median_completion_length": 1375, "min_completion_length": 655, "max_completion_length": 2048, "frac_has_think": 0.76, "frac_has_boxed": 0.76, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 17.02284836769104, "avg_step_time_sec": 25.3219771871289, "gpu_allocated_gb": 32.36479744, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2609.9781107902527, "eta_sec": 2456.2317871515033, "timestamp": "2026-02-15T15:13:19.556502"} |
| {"step": 104, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.509453125, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 52168, "mean_completion_length": 1043.36, "median_completion_length": 1100, "min_completion_length": 399, "max_completion_length": 1821, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 26.373713970184326, "avg_step_time_sec": 25.332090040812126, "gpu_allocated_gb": 32.365656064, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2636.3585152626038, "eta_sec": 2431.8806439179643, "timestamp": "2026-02-15T15:13:43.033273"} |
| {"step": 105, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.96, "mean_correctness_reward": 0.96, "mean_length_penalty": -0.439990234375, "mean_total_reward": 0.96, "std_total_reward": 0.19595917942265426, "total_tokens": 45055, "mean_completion_length": 901.1, "median_completion_length": 736, "min_completion_length": 472, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.12, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 21.90416431427002, "avg_step_time_sec": 25.299443129130772, "gpu_allocated_gb": 32.365307904, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2658.2689821720123, "eta_sec": 2403.4470972674235, "timestamp": "2026-02-15T15:14:05.719545"} |
| {"step": 106, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.572744140625, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 58649, "mean_completion_length": 1172.98, "median_completion_length": 1064, "min_completion_length": 579, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 23.727112293243408, "avg_step_time_sec": 25.28460981935825, "gpu_allocated_gb": 32.365631488, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2682.0032184123993, "eta_sec": 2376.7533230196755, "timestamp": "2026-02-15T15:14:30.788475"} |
| {"step": 107, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.58, "mean_correctness_reward": 0.58, "mean_length_penalty": -0.539775390625, "mean_total_reward": 0.58, "std_total_reward": 0.4935585071701226, "total_tokens": 55273, "mean_completion_length": 1105.46, "median_completion_length": 1105, "min_completion_length": 637, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 25.422482013702393, "avg_step_time_sec": 25.285898344539035, "gpu_allocated_gb": 32.365712384, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2707.43266749382, "eta_sec": 2351.5885460421305, "timestamp": "2026-02-15T15:14:56.289320"} |
| {"step": 108, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.52556640625, "mean_total_reward": 0.74, "std_total_reward": 0.4386342439892262, "total_tokens": 53818, "mean_completion_length": 1076.36, "median_completion_length": 882, "min_completion_length": 420, "max_completion_length": 2048, "frac_has_think": 0.86, "frac_has_boxed": 0.86, "mean_reward_std_per_prompt": 0.0916515138991168, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.239460706710815, "avg_step_time_sec": 25.285468366410996, "gpu_allocated_gb": 32.36565504, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2732.679293870926, "eta_sec": 2326.2630897098115, "timestamp": "2026-02-15T15:15:20.961508"} |
| {"step": 109, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.396923828125, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 40645, "mean_completion_length": 812.9, "median_completion_length": 778, "min_completion_length": 388, "max_completion_length": 1528, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.452012062072754, "avg_step_time_sec": 25.2778219782978, "gpu_allocated_gb": 32.3656192, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2757.137794971466, "eta_sec": 2300.2818000251, "timestamp": "2026-02-15T15:15:41.053851"} |
| {"step": 110, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.96, "mean_correctness_reward": 0.96, "mean_length_penalty": -0.43744140625, "mean_total_reward": 0.96, "std_total_reward": 0.19595917942265426, "total_tokens": 44794, "mean_completion_length": 895.88, "median_completion_length": 857, "min_completion_length": 422, "max_completion_length": 2048, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 17.98546528816223, "avg_step_time_sec": 25.2115278265693, "gpu_allocated_gb": 32.36489984, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2775.1293618679047, "eta_sec": 2269.037504391237, "timestamp": "2026-02-15T15:16:03.036190"} |
| {"step": 111, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.5, "mean_correctness_reward": 0.5, "mean_length_penalty": -0.429921875, "mean_total_reward": 0.5, "std_total_reward": 0.5, "total_tokens": 44024, "mean_completion_length": 880.48, "median_completion_length": 735, "min_completion_length": 420, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.1, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 24.225823402404785, "avg_step_time_sec": 25.20264760653178, "gpu_allocated_gb": 32.365645824, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2799.3612682819366, "eta_sec": 2243.0356369813285, "timestamp": "2026-02-15T15:16:26.833716"} |
| {"step": 112, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.3503125, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 35872, "mean_completion_length": 717.44, "median_completion_length": 669, "min_completion_length": 418, "max_completion_length": 1250, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 23.788440942764282, "avg_step_time_sec": 25.19002076131957, "gpu_allocated_gb": 32.365645824, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2823.1563193798065, "eta_sec": 2216.721826996122, "timestamp": "2026-02-15T15:16:45.102993"} |
| {"step": 113, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.398203125, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 40776, "mean_completion_length": 815.52, "median_completion_length": 669, "min_completion_length": 537, "max_completion_length": 1372, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 15.08491849899292, "avg_step_time_sec": 25.100595077582167, "gpu_allocated_gb": 32.364547584, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2838.247791290283, "eta_sec": 2183.7517717496485, "timestamp": "2026-02-15T15:17:01.831396"} |
| {"step": 114, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.84, "mean_correctness_reward": 0.84, "mean_length_penalty": -0.468125, "mean_total_reward": 0.84, "std_total_reward": 0.3666060555964672, "total_tokens": 47936, "mean_completion_length": 958.72, "median_completion_length": 780, "min_completion_length": 457, "max_completion_length": 2048, "frac_has_think": 0.9, "frac_has_boxed": 0.9, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 17.078352212905884, "avg_step_time_sec": 25.030224526137637, "gpu_allocated_gb": 32.364688384, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2855.3317832946777, "eta_sec": 2152.5993092478366, "timestamp": "2026-02-15T15:17:23.132922"} |
| {"step": 115, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.34400390625, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 35226, "mean_completion_length": 704.52, "median_completion_length": 656, "min_completion_length": 458, "max_completion_length": 1477, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.104089498519897, "avg_step_time_sec": 25.022171178071396, "gpu_allocated_gb": 32.36563968, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2879.4421575069427, "eta_sec": 2126.884550136069, "timestamp": "2026-02-15T15:17:42.486410"} |
| {"step": 116, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.2951953125, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 30228, "mean_completion_length": 604.56, "median_completion_length": 614, "min_completion_length": 293, "max_completion_length": 939, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 16.963985681533813, "avg_step_time_sec": 24.952704061721935, "gpu_allocated_gb": 32.364832256, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2896.4127538204193, "eta_sec": 2096.0271411846425, "timestamp": "2026-02-15T15:17:56.004453"} |
| {"step": 117, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.76, "mean_correctness_reward": 0.76, "mean_length_penalty": -0.384931640625, "mean_total_reward": 0.76, "std_total_reward": 0.4270831300812524, "total_tokens": 39417, "mean_completion_length": 788.34, "median_completion_length": 746, "min_completion_length": 425, "max_completion_length": 2048, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 11.39206600189209, "avg_step_time_sec": 24.836801172321678, "gpu_allocated_gb": 32.364068352, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2907.8117163181305, "eta_sec": 2061.454497302699, "timestamp": "2026-02-15T15:18:15.239265"} |
| {"step": 118, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.494521484375, "mean_total_reward": 0.78, "std_total_reward": 0.4142463035441596, "total_tokens": 50639, "mean_completion_length": 1012.78, "median_completion_length": 1023, "min_completion_length": 479, "max_completion_length": 1847, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.095357418060303, "avg_step_time_sec": 24.830517750675394, "gpu_allocated_gb": 32.365725696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2931.9136698246, "eta_sec": 2036.1024555553822, "timestamp": "2026-02-15T15:18:38.649321"} |
| {"step": 119, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.329130859375, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 33703, "mean_completion_length": 674.06, "median_completion_length": 739, "min_completion_length": 348, "max_completion_length": 1120, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 22.39188289642334, "avg_step_time_sec": 24.810025020807732, "gpu_allocated_gb": 32.36540928, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2954.3121716976166, "eta_sec": 2009.6120266854264, "timestamp": "2026-02-15T15:18:55.305402"} |
| {"step": 120, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.488408203125, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 50013, "mean_completion_length": 1000.26, "median_completion_length": 804, "min_completion_length": 460, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 13.580280303955078, "avg_step_time_sec": 24.71644381483396, "gpu_allocated_gb": 32.364342272, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2967.8986432552338, "eta_sec": 1977.3155051867166, "timestamp": "2026-02-15T15:19:15.875172"} |
| {"step": 121, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.72, "mean_correctness_reward": 0.72, "mean_length_penalty": -0.551171875, "mean_total_reward": 0.72, "std_total_reward": 0.44899888641287294, "total_tokens": 56440, "mean_completion_length": 1128.8, "median_completion_length": 902, "min_completion_length": 548, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.16, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.941832780838013, "avg_step_time_sec": 24.718306533561265, "gpu_allocated_gb": 32.365681664, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2992.8458745479584, "eta_sec": 1952.74621615134, "timestamp": "2026-02-15T15:19:41.269431"} |
| {"step": 122, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.420810546875, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 43091, "mean_completion_length": 861.82, "median_completion_length": 798, "min_completion_length": 446, "max_completion_length": 1685, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.772515058517456, "avg_step_time_sec": 24.718750865733035, "gpu_allocated_gb": 32.365668352, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3017.624593734741, "eta_sec": 1928.0625675271767, "timestamp": "2026-02-15T15:20:02.725784"} |
| {"step": 123, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.372392578125, "mean_total_reward": 0.78, "std_total_reward": 0.4142463035441596, "total_tokens": 38133, "mean_completion_length": 762.66, "median_completion_length": 645, "min_completion_length": 214, "max_completion_length": 1398, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 19.834197759628296, "avg_step_time_sec": 24.67903905186227, "gpu_allocated_gb": 32.365136384, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3037.464989423752, "eta_sec": 1900.2860069933947, "timestamp": "2026-02-15T15:20:20.476926"} |
| {"step": 124, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.2992578125, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 30644, "mean_completion_length": 612.88, "median_completion_length": 612, "min_completion_length": 307, "max_completion_length": 1140, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 16.597280502319336, "avg_step_time_sec": 24.613863579688534, "gpu_allocated_gb": 32.364716544, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3054.068367242813, "eta_sec": 1870.6536320563287, "timestamp": "2026-02-15T15:20:35.379159"} |
| {"step": 125, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.565703125, "mean_total_reward": 0.4, "std_total_reward": 0.4898979485566356, "total_tokens": 57928, "mean_completion_length": 1158.56, "median_completion_length": 805, "min_completion_length": 419, "max_completion_length": 2048, "frac_has_think": 0.7, "frac_has_boxed": 0.7, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 13.940901756286621, "avg_step_time_sec": 24.52847988510132, "gpu_allocated_gb": 32.364361728, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3068.0156588554382, "eta_sec": 1839.635991382599, "timestamp": "2026-02-15T15:20:56.862185"} |
| {"step": 126, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.88, "mean_correctness_reward": 0.88, "mean_length_penalty": -0.43392578125, "mean_total_reward": 0.88, "std_total_reward": 0.32496153618543844, "total_tokens": 44434, "mean_completion_length": 888.68, "median_completion_length": 769, "min_completion_length": 335, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.16, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.451818227767944, "avg_step_time_sec": 24.535807967185974, "gpu_allocated_gb": 32.365696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3093.4734194278717, "eta_sec": 1815.649789571762, "timestamp": "2026-02-15T15:21:21.056982"} |
| {"step": 127, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.4008203125, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 41044, "mean_completion_length": 820.88, "median_completion_length": 819, "min_completion_length": 486, "max_completion_length": 1433, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 23.88256549835205, "avg_step_time_sec": 24.530664325699092, "gpu_allocated_gb": 32.365637632, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3117.3615651130676, "eta_sec": 1790.7384957760337, "timestamp": "2026-02-15T15:21:40.770218"} |
| {"step": 128, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.54, "mean_correctness_reward": 0.54, "mean_length_penalty": -0.46271484375, "mean_total_reward": 0.54, "std_total_reward": 0.4983974317750845, "total_tokens": 47382, "mean_completion_length": 947.64, "median_completion_length": 726, "min_completion_length": 408, "max_completion_length": 2048, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0916515138991168, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 17.471781969070435, "avg_step_time_sec": 24.47551680728793, "gpu_allocated_gb": 32.364763136, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3134.8396866321564, "eta_sec": 1762.237210124731, "timestamp": "2026-02-15T15:22:02.583566"} |
| {"step": 129, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.82, "mean_correctness_reward": 0.82, "mean_length_penalty": -0.45244140625, "mean_total_reward": 0.82, "std_total_reward": 0.3841874542459709, "total_tokens": 46330, "mean_completion_length": 926.6, "median_completion_length": 694, "min_completion_length": 408, "max_completion_length": 2048, "frac_has_think": 0.82, "frac_has_boxed": 0.82, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.095845937728882, "avg_step_time_sec": 24.47257362225259, "gpu_allocated_gb": 32.365645824, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3158.941735267639, "eta_sec": 1737.5527271799338, "timestamp": "2026-02-15T15:22:26.713146"} |
| {"step": 130, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.512431640625, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 52473, "mean_completion_length": 1049.46, "median_completion_length": 1079, "min_completion_length": 370, "max_completion_length": 1782, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.068299055099487, "avg_step_time_sec": 24.469463817889874, "gpu_allocated_gb": 32.365640704, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3183.016888856888, "eta_sec": 1712.8624672522913, "timestamp": "2026-02-15T15:22:49.544809"} |
| {"step": 131, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.3975, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 40704, "mean_completion_length": 814.08, "median_completion_length": 808, "min_completion_length": 501, "max_completion_length": 1180, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 22.145960569381714, "avg_step_time_sec": 24.451727151870728, "gpu_allocated_gb": 32.365272576, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3205.169408798218, "eta_sec": 1687.1691734790802, "timestamp": "2026-02-15T15:23:07.035226"} |
| {"step": 132, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.96, "mean_correctness_reward": 0.96, "mean_length_penalty": -0.451884765625, "mean_total_reward": 0.96, "std_total_reward": 0.19595917942265426, "total_tokens": 46273, "mean_completion_length": 925.46, "median_completion_length": 977, "min_completion_length": 170, "max_completion_length": 2033, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.12, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 14.559662103652954, "avg_step_time_sec": 24.376787265141804, "gpu_allocated_gb": 32.364390912, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3219.7344205379486, "eta_sec": 1657.6215340296426, "timestamp": "2026-02-15T15:23:27.072105"} |
| {"step": 133, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.84, "mean_correctness_reward": 0.84, "mean_length_penalty": -0.381044921875, "mean_total_reward": 0.84, "std_total_reward": 0.3666060555964672, "total_tokens": 39019, "mean_completion_length": 780.38, "median_completion_length": 595, "min_completion_length": 390, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.1516515138991168, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 23.58193325996399, "avg_step_time_sec": 24.370810919238213, "gpu_allocated_gb": 32.365629952, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3243.3229682445526, "eta_sec": 1632.8443315889604, "timestamp": "2026-02-15T15:23:50.395177"} |
| {"step": 134, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.62, "mean_correctness_reward": 0.62, "mean_length_penalty": -0.634599609375, "mean_total_reward": 0.62, "std_total_reward": 0.4853864439804639, "total_tokens": 64983, "mean_completion_length": 1299.66, "median_completion_length": 1016, "min_completion_length": 709, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.82, "mean_reward_std_per_prompt": 0.26, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 23.374454021453857, "avg_step_time_sec": 24.363375420001017, "gpu_allocated_gb": 32.365641728, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3266.7042846679688, "eta_sec": 1607.9827777200671, "timestamp": "2026-02-15T15:24:16.561138"} |
| {"step": 135, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.76, "mean_correctness_reward": 0.76, "mean_length_penalty": -0.469638671875, "mean_total_reward": 0.76, "std_total_reward": 0.4270831300812524, "total_tokens": 48091, "mean_completion_length": 961.82, "median_completion_length": 867, "min_completion_length": 612, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.12, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 26.22987699508667, "avg_step_time_sec": 24.37720135759424, "gpu_allocated_gb": 32.365657088, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3292.9406311511993, "eta_sec": 1584.5180882436257, "timestamp": "2026-02-15T15:24:40.512845"} |
| {"step": 136, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.39025390625, "mean_total_reward": 0.4, "std_total_reward": 0.4898979485566356, "total_tokens": 39962, "mean_completion_length": 799.24, "median_completion_length": 820, "min_completion_length": 451, "max_completion_length": 1496, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 23.754759311676025, "avg_step_time_sec": 24.372624577844846, "gpu_allocated_gb": 32.365615104, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3316.7018599510193, "eta_sec": 1559.8479729820701, "timestamp": "2026-02-15T15:25:00.314340"} |
| {"step": 137, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.363486328125, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 37221, "mean_completion_length": 744.42, "median_completion_length": 602, "min_completion_length": 413, "max_completion_length": 1785, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 17.664690017700195, "avg_step_time_sec": 24.323661551858386, "gpu_allocated_gb": 32.364862976, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3334.3727850914, "eta_sec": 1532.3906777670784, "timestamp": "2026-02-15T15:25:19.622558"} |
| {"step": 138, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.68, "mean_correctness_reward": 0.68, "mean_length_penalty": -0.458583984375, "mean_total_reward": 0.68, "std_total_reward": 0.466476151587624, "total_tokens": 46959, "mean_completion_length": 939.18, "median_completion_length": 739, "min_completion_length": 425, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.16, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 20.933507442474365, "avg_step_time_sec": 24.299095217732415, "gpu_allocated_gb": 32.365279232, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3355.312567949295, "eta_sec": 1506.5439034994097, "timestamp": "2026-02-15T15:25:43.107628"} |
| {"step": 139, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.464306640625, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 47545, "mean_completion_length": 950.9, "median_completion_length": 723, "min_completion_length": 481, "max_completion_length": 2048, "frac_has_think": 0.82, "frac_has_boxed": 0.82, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.18170428276062, "avg_step_time_sec": 24.2982506786319, "gpu_allocated_gb": 32.365650944, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3379.5004873275757, "eta_sec": 1482.1932913965459, "timestamp": "2026-02-15T15:26:07.403135"} |
| {"step": 140, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.3846875, "mean_total_reward": 0.78, "std_total_reward": 0.4142463035441596, "total_tokens": 39392, "mean_completion_length": 787.84, "median_completion_length": 725, "min_completion_length": 222, "max_completion_length": 1432, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.4332332611084, "avg_step_time_sec": 24.299214839935303, "gpu_allocated_gb": 32.365678592, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3403.9399976730347, "eta_sec": 1457.9528903961182, "timestamp": "2026-02-15T15:26:27.287300"} |
| {"step": 141, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.92, "mean_correctness_reward": 0.92, "mean_length_penalty": -0.498681640625, "mean_total_reward": 0.92, "std_total_reward": 0.2712931993250107, "total_tokens": 51065, "mean_completion_length": 1021.3, "median_completion_length": 927, "min_completion_length": 131, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.09797958971132711, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 17.085137367248535, "avg_step_time_sec": 24.24805116991625, "gpu_allocated_gb": 32.364782592, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3421.0321140289307, "eta_sec": 1430.6350190250587, "timestamp": "2026-02-15T15:26:49.674652"} |
| {"step": 142, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.398876953125, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 40845, "mean_completion_length": 816.9, "median_completion_length": 836, "min_completion_length": 459, "max_completion_length": 1273, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.953808307647705, "avg_step_time_sec": 24.253021290604497, "gpu_allocated_gb": 32.365637632, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3445.9924376010895, "eta_sec": 1406.6752348550608, "timestamp": "2026-02-15T15:27:08.415653"} |
| {"step": 143, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.64, "mean_correctness_reward": 0.64, "mean_length_penalty": -0.50974609375, "mean_total_reward": 0.64, "std_total_reward": 0.48, "total_tokens": 52198, "mean_completion_length": 1043.96, "median_completion_length": 780, "min_completion_length": 451, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 15.550028085708618, "avg_step_time_sec": 24.192161198262568, "gpu_allocated_gb": 32.364539904, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3461.54972243309, "eta_sec": 1378.9531883009663, "timestamp": "2026-02-15T15:27:29.766944"} |
| {"step": 144, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.351142578125, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 35957, "mean_completion_length": 719.14, "median_completion_length": 648, "min_completion_length": 325, "max_completion_length": 1853, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 24.542055130004883, "avg_step_time_sec": 24.194591017233, "gpu_allocated_gb": 32.365647872, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3486.097778081894, "eta_sec": 1354.8970969650481, "timestamp": "2026-02-15T15:27:51.468266"} |
| {"step": 145, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.413310546875, "mean_total_reward": 0.98, "std_total_reward": 0.14, "total_tokens": 42323, "mean_completion_length": 846.46, "median_completion_length": 839, "min_completion_length": 387, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 21.340415477752686, "avg_step_time_sec": 24.174907047995205, "gpu_allocated_gb": 32.365357056, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3507.4443089962006, "eta_sec": 1329.6198876397364, "timestamp": "2026-02-15T15:28:14.443966"} |
| {"step": 146, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.406171875, "mean_total_reward": 0.98, "std_total_reward": 0.14, "total_tokens": 41592, "mean_completion_length": 831.84, "median_completion_length": 809, "min_completion_length": 309, "max_completion_length": 1644, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 23.36310076713562, "avg_step_time_sec": 24.169346731003017, "gpu_allocated_gb": 32.365657088, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3530.813656568527, "eta_sec": 1305.144723474163, "timestamp": "2026-02-15T15:28:35.437478"} |
| {"step": 147, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.40865234375, "mean_total_reward": 0.98, "std_total_reward": 0.14, "total_tokens": 41846, "mean_completion_length": 836.92, "median_completion_length": 821, "min_completion_length": 301, "max_completion_length": 1523, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 19.204582929611206, "avg_step_time_sec": 24.13557282759219, "gpu_allocated_gb": 32.365074944, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3550.024359703064, "eta_sec": 1279.185359862386, "timestamp": "2026-02-15T15:28:54.119174"} |
| {"step": 148, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.52, "mean_correctness_reward": 0.52, "mean_length_penalty": -0.56630859375, "mean_total_reward": 0.52, "std_total_reward": 0.49959983987187184, "total_tokens": 57990, "mean_completion_length": 1159.8, "median_completion_length": 1191, "min_completion_length": 601, "max_completion_length": 1625, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.1516515138991168, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 18.230257749557495, "avg_step_time_sec": 24.095672050037898, "gpu_allocated_gb": 32.36491008, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3568.2604932785034, "eta_sec": 1252.9749466019707, "timestamp": "2026-02-15T15:29:14.788774"} |
| {"step": 149, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.317314453125, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 32493, "mean_completion_length": 649.86, "median_completion_length": 666, "min_completion_length": 269, "max_completion_length": 1185, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 21.190475940704346, "avg_step_time_sec": 24.076174089572575, "gpu_allocated_gb": 32.365065728, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3589.456933259964, "eta_sec": 1227.8848785682014, "timestamp": "2026-02-15T15:29:30.721729"} |
| {"step": 150, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.92, "mean_correctness_reward": 0.92, "mean_length_penalty": -0.440322265625, "mean_total_reward": 0.92, "std_total_reward": 0.2712931993250107, "total_tokens": 45089, "mean_completion_length": 901.78, "median_completion_length": 641, "min_completion_length": 392, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.09797958971132711, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 13.936412811279297, "avg_step_time_sec": 24.00857568105062, "gpu_allocated_gb": 32.364396032, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3603.4004797935486, "eta_sec": 1200.428784052531, "timestamp": "2026-02-15T15:29:51.667161"} |
| {"step": 151, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.32927734375, "mean_total_reward": 0.98, "std_total_reward": 0.14, "total_tokens": 33718, "mean_completion_length": 674.36, "median_completion_length": 656, "min_completion_length": 262, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 23.963045358657837, "avg_step_time_sec": 24.00827415573676, "gpu_allocated_gb": 32.365632512, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3627.961492061615, "eta_sec": 1176.4054336311012, "timestamp": "2026-02-15T15:30:14.231587"} |
| {"step": 152, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.54, "mean_correctness_reward": 0.54, "mean_length_penalty": -0.703095703125, "mean_total_reward": 0.54, "std_total_reward": 0.4983974317750845, "total_tokens": 71997, "mean_completion_length": 1439.94, "median_completion_length": 1500, "min_completion_length": 562, "max_completion_length": 2048, "frac_has_think": 0.84, "frac_has_boxed": 0.84, "mean_reward_std_per_prompt": 0.1579795897113271, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 23.326800107955933, "avg_step_time_sec": 24.003790773843463, "gpu_allocated_gb": 32.365696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3651.295053243637, "eta_sec": 1152.1819571444862, "timestamp": "2026-02-15T15:30:41.163435"} |
| {"step": 153, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.4352734375, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 44572, "mean_completion_length": 891.44, "median_completion_length": 878, "min_completion_length": 624, "max_completion_length": 1350, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 26.203153133392334, "avg_step_time_sec": 24.018165691226137, "gpu_allocated_gb": 32.365645824, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3677.5040237903595, "eta_sec": 1128.8537874876286, "timestamp": "2026-02-15T15:31:00.815205"} |
| {"step": 154, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.56869140625, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 58234, "mean_completion_length": 1164.68, "median_completion_length": 1042, "min_completion_length": 516, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 16.91133761405945, "avg_step_time_sec": 23.97201745695882, "gpu_allocated_gb": 32.36471296, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3694.421368598938, "eta_sec": 1102.7128030201056, "timestamp": "2026-02-15T15:31:23.691769"} |
| {"step": 155, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.464521484375, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 47567, "mean_completion_length": 951.34, "median_completion_length": 692, "min_completion_length": 500, "max_completion_length": 2048, "frac_has_think": 0.88, "frac_has_boxed": 0.88, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.5807101726532, "avg_step_time_sec": 23.98239611964072, "gpu_allocated_gb": 32.365646848, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3720.008092880249, "eta_sec": 1079.2078253838324, "timestamp": "2026-02-15T15:31:47.863938"} |
| {"step": 156, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.311494140625, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 31897, "mean_completion_length": 637.94, "median_completion_length": 622, "min_completion_length": 324, "max_completion_length": 1210, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.18597102165222, "avg_step_time_sec": 23.983701086961307, "gpu_allocated_gb": 32.365668352, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3744.2005014419556, "eta_sec": 1055.2828478262975, "timestamp": "2026-02-15T15:32:05.652075"} |
| {"step": 157, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.557041015625, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 57041, "mean_completion_length": 1140.82, "median_completion_length": 881, "min_completion_length": 481, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 14.415817975997925, "avg_step_time_sec": 23.92275915631823, "gpu_allocated_gb": 32.364476928, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3758.6229219436646, "eta_sec": 1028.6786437216838, "timestamp": "2026-02-15T15:32:27.259025"} |
| {"step": 158, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.3403125, "mean_total_reward": 0.98, "std_total_reward": 0.14, "total_tokens": 34848, "mean_completion_length": 696.96, "median_completion_length": 662, "min_completion_length": 408, "max_completion_length": 1285, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.175652027130127, "avg_step_time_sec": 23.930688858032227, "gpu_allocated_gb": 32.36568064, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3783.8051080703735, "eta_sec": 1005.0889320373535, "timestamp": "2026-02-15T15:32:46.423315"} |
| {"step": 159, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.499482421875, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 51147, "mean_completion_length": 1022.94, "median_completion_length": 962, "min_completion_length": 561, "max_completion_length": 1871, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 15.98077940940857, "avg_step_time_sec": 23.88068942753774, "gpu_allocated_gb": 32.364606464, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3799.7922234535217, "eta_sec": 979.1082665290473, "timestamp": "2026-02-15T15:33:06.789791"} |
| {"step": 160, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.553037109375, "mean_total_reward": 0.74, "std_total_reward": 0.4386342439892262, "total_tokens": 56631, "mean_completion_length": 1132.62, "median_completion_length": 957, "min_completion_length": 511, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.0916515138991168, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 22.57667088508606, "avg_step_time_sec": 23.872539311647415, "gpu_allocated_gb": 32.365398528, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3822.376928806305, "eta_sec": 954.9015724658966, "timestamp": "2026-02-15T15:33:30.973379"} |
| {"step": 161, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.437724609375, "mean_total_reward": 0.78, "std_total_reward": 0.4142463035441596, "total_tokens": 44823, "mean_completion_length": 896.46, "median_completion_length": 872, "min_completion_length": 418, "max_completion_length": 1617, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.156012535095215, "avg_step_time_sec": 23.880511195022866, "gpu_allocated_gb": 32.365683712, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3847.5393154621124, "eta_sec": 931.3399366058918, "timestamp": "2026-02-15T15:33:52.984034"} |
| {"step": 162, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.88, "mean_correctness_reward": 0.88, "mean_length_penalty": -0.329658203125, "mean_total_reward": 0.88, "std_total_reward": 0.32496153618543844, "total_tokens": 33757, "mean_completion_length": 675.14, "median_completion_length": 662, "min_completion_length": 387, "max_completion_length": 1205, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.09797958971132711, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 19.88985800743103, "avg_step_time_sec": 23.855877533371064, "gpu_allocated_gb": 32.365016064, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3867.4368555545807, "eta_sec": 906.5233462681005, "timestamp": "2026-02-15T15:34:08.955861"} |
| {"step": 163, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.491572265625, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 50337, "mean_completion_length": 1006.74, "median_completion_length": 985, "min_completion_length": 627, "max_completion_length": 2048, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 14.50003695487976, "avg_step_time_sec": 23.798479738410997, "gpu_allocated_gb": 32.364481024, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3881.9431331157684, "eta_sec": 880.5437503212069, "timestamp": "2026-02-15T15:34:29.605312"} |
| {"step": 164, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.346748046875, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 35507, "mean_completion_length": 710.14, "median_completion_length": 669, "min_completion_length": 489, "max_completion_length": 1127, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 23.967074871063232, "avg_step_time_sec": 23.799507757512533, "gpu_allocated_gb": 32.365625344, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3905.916805744171, "eta_sec": 856.7822792704512, "timestamp": "2026-02-15T15:34:47.195614"} |
| {"step": 165, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.3217578125, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 32948, "mean_completion_length": 658.96, "median_completion_length": 672, "min_completion_length": 388, "max_completion_length": 1038, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 14.552917242050171, "avg_step_time_sec": 23.74346781499458, "gpu_allocated_gb": 32.364393984, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3920.4789073467255, "eta_sec": 831.0213735248103, "timestamp": "2026-02-15T15:35:00.905439"} |
| {"step": 166, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.76, "mean_correctness_reward": 0.76, "mean_length_penalty": -0.586982421875, "mean_total_reward": 0.76, "std_total_reward": 0.4270831300812524, "total_tokens": 60107, "mean_completion_length": 1202.14, "median_completion_length": 1128, "min_completion_length": 278, "max_completion_length": 2048, "frac_has_think": 0.9, "frac_has_boxed": 0.9, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 12.627315282821655, "avg_step_time_sec": 23.676503040704382, "gpu_allocated_gb": 32.364210176, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3933.1123366355896, "eta_sec": 805.001103383949, "timestamp": "2026-02-15T15:35:22.071172"} |
| {"step": 167, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.363427734375, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 37215, "mean_completion_length": 744.3, "median_completion_length": 677, "min_completion_length": 368, "max_completion_length": 1578, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.475756406784058, "avg_step_time_sec": 23.687277012956358, "gpu_allocated_gb": 32.365681664, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3958.594392299652, "eta_sec": 781.6801414275598, "timestamp": "2026-02-15T15:35:42.707451"} |
| {"step": 168, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.36, "mean_correctness_reward": 0.36, "mean_length_penalty": -0.5694921875, "mean_total_reward": 0.36, "std_total_reward": 0.48, "total_tokens": 58316, "mean_completion_length": 1166.32, "median_completion_length": 1062, "min_completion_length": 504, "max_completion_length": 2048, "frac_has_think": 0.9, "frac_has_boxed": 0.9, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 18.63668656349182, "avg_step_time_sec": 23.657213974566687, "gpu_allocated_gb": 32.365020672, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 3977.237099647522, "eta_sec": 757.030847186134, "timestamp": "2026-02-15T15:36:06.453609"} |
| {"step": 169, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.38962890625, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 39898, "mean_completion_length": 797.96, "median_completion_length": 800, "min_completion_length": 545, "max_completion_length": 1523, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.531309604644775, "avg_step_time_sec": 23.668303297821588, "gpu_allocated_gb": 32.365631488, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4002.774299621582, "eta_sec": 733.7174022324692, "timestamp": "2026-02-15T15:36:26.484572"} |
| {"step": 170, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.393203125, "mean_total_reward": 0.98, "std_total_reward": 0.14, "total_tokens": 40264, "mean_completion_length": 805.28, "median_completion_length": 801, "min_completion_length": 438, "max_completion_length": 1208, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 17.765581607818604, "avg_step_time_sec": 23.63358140552745, "gpu_allocated_gb": 32.364883456, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4020.545969247818, "eta_sec": 709.0074421658235, "timestamp": "2026-02-15T15:36:42.571624"} |
| {"step": 171, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.335283203125, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 34333, "mean_completion_length": 686.66, "median_completion_length": 720, "min_completion_length": 439, "max_completion_length": 1355, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 14.935601711273193, "avg_step_time_sec": 23.58271602719848, "gpu_allocated_gb": 32.36445696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4035.487459421158, "eta_sec": 683.8987647887559, "timestamp": "2026-02-15T15:36:57.826991"} |
| {"step": 172, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.84, "mean_correctness_reward": 0.84, "mean_length_penalty": -0.5056640625, "mean_total_reward": 0.84, "std_total_reward": 0.3666060555964672, "total_tokens": 51780, "mean_completion_length": 1035.6, "median_completion_length": 1013, "min_completion_length": 518, "max_completion_length": 1887, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.08, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 16.04830050468445, "avg_step_time_sec": 23.538911285788515, "gpu_allocated_gb": 32.364731392, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4051.541830778122, "eta_sec": 659.0895160020784, "timestamp": "2026-02-15T15:37:18.995136"} |
| {"step": 173, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.360029296875, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 36867, "mean_completion_length": 737.34, "median_completion_length": 715, "min_completion_length": 287, "max_completion_length": 1493, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 23.074387073516846, "avg_step_time_sec": 23.53622617473492, "gpu_allocated_gb": 32.365416448, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4074.6220285892487, "eta_sec": 635.4781067178428, "timestamp": "2026-02-15T15:37:38.034629"} |
| {"step": 174, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.4360546875, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 44652, "mean_completion_length": 893.04, "median_completion_length": 870, "min_completion_length": 336, "max_completion_length": 1581, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 17.313328981399536, "avg_step_time_sec": 23.50046239776173, "gpu_allocated_gb": 32.364845056, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4091.9415266513824, "eta_sec": 611.0120223418049, "timestamp": "2026-02-15T15:37:56.554725"} |
| {"step": 175, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.419736328125, "mean_total_reward": 0.78, "std_total_reward": 0.4142463035441596, "total_tokens": 42981, "mean_completion_length": 859.62, "median_completion_length": 859, "min_completion_length": 410, "max_completion_length": 1498, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 18.86185073852539, "avg_step_time_sec": 23.473956045423236, "gpu_allocated_gb": 32.364983296, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4110.808932542801, "eta_sec": 586.848901135581, "timestamp": "2026-02-15T15:38:15.326368"} |
| {"step": 176, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.374775390625, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 38377, "mean_completion_length": 767.54, "median_completion_length": 782, "min_completion_length": 369, "max_completion_length": 1297, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 18.526092052459717, "avg_step_time_sec": 23.445843181826852, "gpu_allocated_gb": 32.364870656, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4129.340816736221, "eta_sec": 562.7002363638444, "timestamp": "2026-02-15T15:38:31.676821"} |
| {"step": 177, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.378974609375, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 38807, "mean_completion_length": 776.14, "median_completion_length": 755, "min_completion_length": 530, "max_completion_length": 1320, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 15.427118301391602, "avg_step_time_sec": 23.40053965142891, "gpu_allocated_gb": 32.364564992, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4144.774101257324, "eta_sec": 538.212411982865, "timestamp": "2026-02-15T15:38:47.247099"} |
| {"step": 178, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.355234375, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 36376, "mean_completion_length": 727.52, "median_completion_length": 731, "min_completion_length": 431, "max_completion_length": 1323, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 15.675445318222046, "avg_step_time_sec": 23.35714024506258, "gpu_allocated_gb": 32.364624896, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4160.455217599869, "eta_sec": 513.8570853913768, "timestamp": "2026-02-15T15:39:02.785516"} |
| {"step": 179, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.49626953125, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 50818, "mean_completion_length": 1016.36, "median_completion_length": 758, "min_completion_length": 411, "max_completion_length": 2048, "frac_has_think": 0.84, "frac_has_boxed": 0.84, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 15.705410718917847, "avg_step_time_sec": 23.31439315273775, "gpu_allocated_gb": 32.364631552, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4176.166404008865, "eta_sec": 489.60225620749276, "timestamp": "2026-02-15T15:39:24.860010"} |
| {"step": 180, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.86, "mean_correctness_reward": 0.86, "mean_length_penalty": -0.54662109375, "mean_total_reward": 0.86, "std_total_reward": 0.34698703145794946, "total_tokens": 55974, "mean_completion_length": 1119.48, "median_completion_length": 887, "min_completion_length": 460, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.1579795897113271, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.331791877746582, "avg_step_time_sec": 23.325600923432244, "gpu_allocated_gb": 32.365698048, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4201.504014015198, "eta_sec": 466.5120184686449, "timestamp": "2026-02-15T15:39:50.042583"} |
| {"step": 181, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.7, "mean_correctness_reward": 0.7, "mean_length_penalty": -0.54013671875, "mean_total_reward": 0.7, "std_total_reward": 0.458257569495584, "total_tokens": 55310, "mean_completion_length": 1106.2, "median_completion_length": 853, "min_completion_length": 475, "max_completion_length": 2048, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.17165151389911681, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.896844148635864, "avg_step_time_sec": 23.33428182522895, "gpu_allocated_gb": 32.365654016, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4226.406719446182, "eta_sec": 443.35135467935004, "timestamp": "2026-02-15T15:40:14.812173"} |
| {"step": 182, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.32921875, "mean_total_reward": 0.98, "std_total_reward": 0.14, "total_tokens": 33712, "mean_completion_length": 674.24, "median_completion_length": 649, "min_completion_length": 328, "max_completion_length": 1235, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.68673586845398, "avg_step_time_sec": 23.341712891400515, "gpu_allocated_gb": 32.365630464, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4251.1006054878235, "eta_sec": 420.15083204520926, "timestamp": "2026-02-15T15:40:33.374989"} |
| {"step": 183, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.461181640625, "mean_total_reward": 0.74, "std_total_reward": 0.4386342439892262, "total_tokens": 47225, "mean_completion_length": 944.5, "median_completion_length": 699, "min_completion_length": 452, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.0916515138991168, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 15.214890718460083, "avg_step_time_sec": 23.297304027067508, "gpu_allocated_gb": 32.364482048, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4266.3222460746765, "eta_sec": 396.05416846014765, "timestamp": "2026-02-15T15:40:54.130383"} |
| {"step": 184, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.58, "mean_correctness_reward": 0.58, "mean_length_penalty": -0.508212890625, "mean_total_reward": 0.58, "std_total_reward": 0.4935585071701226, "total_tokens": 52041, "mean_completion_length": 1040.82, "median_completion_length": 758, "min_completion_length": 445, "max_completion_length": 2048, "frac_has_think": 0.86, "frac_has_boxed": 0.86, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 24.099818229675293, "avg_step_time_sec": 23.30166551729907, "gpu_allocated_gb": 32.365635584, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4290.428211688995, "eta_sec": 372.82664827678514, "timestamp": "2026-02-15T15:41:18.724296"} |
| {"step": 185, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.54, "mean_correctness_reward": 0.54, "mean_length_penalty": -0.43486328125, "mean_total_reward": 0.54, "std_total_reward": 0.4983974317750845, "total_tokens": 44530, "mean_completion_length": 890.6, "median_completion_length": 678, "min_completion_length": 446, "max_completion_length": 2019, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0916515138991168, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 24.690062284469604, "avg_step_time_sec": 23.309170364689184, "gpu_allocated_gb": 32.365664256, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4315.124134063721, "eta_sec": 349.63755547033776, "timestamp": "2026-02-15T15:41:42.509994"} |
| {"step": 186, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.36283203125, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 37154, "mean_completion_length": 743.08, "median_completion_length": 706, "min_completion_length": 514, "max_completion_length": 1537, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.310137510299683, "avg_step_time_sec": 23.314551908482787, "gpu_allocated_gb": 32.365655552, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4339.440948724747, "eta_sec": 326.403726718759, "timestamp": "2026-02-15T15:42:03.087840"} |
| {"step": 187, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.32501953125, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 33282, "mean_completion_length": 665.64, "median_completion_length": 641, "min_completion_length": 375, "max_completion_length": 1089, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 17.83823847770691, "avg_step_time_sec": 23.28526680992249, "gpu_allocated_gb": 32.364938752, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4357.285480976105, "eta_sec": 302.7084685289924, "timestamp": "2026-02-15T15:42:18.206793"} |
| {"step": 188, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.534326171875, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 54715, "mean_completion_length": 1094.3, "median_completion_length": 990, "min_completion_length": 469, "max_completion_length": 2048, "frac_has_think": 0.92, "frac_has_boxed": 0.92, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 13.613678455352783, "avg_step_time_sec": 23.233822191015204, "gpu_allocated_gb": 32.3643648, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4370.906333208084, "eta_sec": 278.80586629218243, "timestamp": "2026-02-15T15:42:39.383882"} |
| {"step": 189, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.36193359375, "mean_total_reward": 0.98, "std_total_reward": 0.14, "total_tokens": 37062, "mean_completion_length": 741.24, "median_completion_length": 623, "min_completion_length": 354, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.836175203323364, "avg_step_time_sec": 23.242300249281385, "gpu_allocated_gb": 32.365657088, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4395.747663259506, "eta_sec": 255.66530274209524, "timestamp": "2026-02-15T15:43:03.017737"} |
| {"step": 190, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.377197265625, "mean_total_reward": 0.98, "std_total_reward": 0.14, "total_tokens": 38625, "mean_completion_length": 772.5, "median_completion_length": 857, "min_completion_length": 275, "max_completion_length": 1228, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 23.431544303894043, "avg_step_time_sec": 23.24329627062145, "gpu_allocated_gb": 32.365618176, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4419.185643672943, "eta_sec": 232.43296270621448, "timestamp": "2026-02-15T15:43:21.211128"} |
| {"step": 191, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.53943359375, "mean_total_reward": 0.4, "std_total_reward": 0.4898979485566356, "total_tokens": 55238, "mean_completion_length": 1104.76, "median_completion_length": 969, "min_completion_length": 413, "max_completion_length": 2048, "frac_has_think": 0.9, "frac_has_boxed": 0.9, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 14.996856451034546, "avg_step_time_sec": 23.20012119303199, "gpu_allocated_gb": 32.364491776, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4434.1900181770325, "eta_sec": 208.8010907372879, "timestamp": "2026-02-15T15:43:42.872552"} |
| {"step": 192, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.364501953125, "mean_total_reward": 0.6, "std_total_reward": 0.4898979485566356, "total_tokens": 37325, "mean_completion_length": 746.5, "median_completion_length": 640, "min_completion_length": 374, "max_completion_length": 1676, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 25.469441652297974, "avg_step_time_sec": 23.211940570424, "gpu_allocated_gb": 32.365751296, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4459.6654760837555, "eta_sec": 185.695524563392, "timestamp": "2026-02-15T15:44:04.300000"} |
| {"step": 193, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.517587890625, "mean_total_reward": 0.8, "std_total_reward": 0.4, "total_tokens": 53001, "mean_completion_length": 1060.02, "median_completion_length": 849, "min_completion_length": 520, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 19.309236526489258, "avg_step_time_sec": 23.191719305947654, "gpu_allocated_gb": 32.365103616, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4478.980659008026, "eta_sec": 162.34203514163357, "timestamp": "2026-02-15T15:44:27.812928"} |
| {"step": 194, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.58, "mean_correctness_reward": 0.58, "mean_length_penalty": -0.38423828125, "mean_total_reward": 0.58, "std_total_reward": 0.4935585071701226, "total_tokens": 39346, "mean_completion_length": 786.92, "median_completion_length": 741, "min_completion_length": 476, "max_completion_length": 1615, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 25.471590280532837, "avg_step_time_sec": 23.203471218187783, "gpu_allocated_gb": 32.365712384, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4504.458561897278, "eta_sec": 139.2208273091267, "timestamp": "2026-02-15T15:44:48.616866"} |
| {"step": 195, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.66, "mean_correctness_reward": 0.66, "mean_length_penalty": -0.434404296875, "mean_total_reward": 0.66, "std_total_reward": 0.4737087712930804, "total_tokens": 44483, "mean_completion_length": 889.66, "median_completion_length": 640, "min_completion_length": 355, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.0916515138991168, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 18.67878794670105, "avg_step_time_sec": 23.18026771423144, "gpu_allocated_gb": 32.365039104, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4523.144848585129, "eta_sec": 115.9013385711572, "timestamp": "2026-02-15T15:45:10.633169"} |
| {"step": 196, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.9, "mean_correctness_reward": 0.9, "mean_length_penalty": -0.50732421875, "mean_total_reward": 0.9, "std_total_reward": 0.3, "total_tokens": 51950, "mean_completion_length": 1039, "median_completion_length": 855, "min_completion_length": 519, "max_completion_length": 2048, "frac_has_think": 0.92, "frac_has_boxed": 0.92, "mean_reward_std_per_prompt": 0.1, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 23.758422136306763, "avg_step_time_sec": 23.18321748169101, "gpu_allocated_gb": 32.365627392, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4546.909566402435, "eta_sec": 92.73286992676404, "timestamp": "2026-02-15T15:45:35.624665"} |
| {"step": 197, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.64, "mean_correctness_reward": 0.64, "mean_length_penalty": -0.498095703125, "mean_total_reward": 0.64, "std_total_reward": 0.48, "total_tokens": 51005, "mean_completion_length": 1020.1, "median_completion_length": 831, "min_completion_length": 372, "max_completion_length": 2048, "frac_has_think": 0.88, "frac_has_boxed": 0.88, "mean_reward_std_per_prompt": 0.1516515138991168, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.0209858417511, "avg_step_time_sec": 23.19254625509233, "gpu_allocated_gb": 32.365645824, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4571.9378209114075, "eta_sec": 69.57763876527699, "timestamp": "2026-02-15T15:46:00.390696"} |
| {"step": 198, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.26548828125, "mean_total_reward": 1.0, "std_total_reward": 0.0, "total_tokens": 27186, "mean_completion_length": 543.72, "median_completion_length": 541, "min_completion_length": 310, "max_completion_length": 888, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.672580242156982, "avg_step_time_sec": 23.205071679269423, "gpu_allocated_gb": 32.365825024, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4597.616838693619, "eta_sec": 46.410143358538846, "timestamp": "2026-02-15T15:46:16.730626"} |
| {"step": 199, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.59376953125, "mean_total_reward": 0.4, "std_total_reward": 0.4898979485566356, "total_tokens": 60802, "mean_completion_length": 1216.04, "median_completion_length": 866, "min_completion_length": 506, "max_completion_length": 2048, "frac_has_think": 0.76, "frac_has_boxed": 0.76, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 10.783756494522095, "avg_step_time_sec": 23.14265300999934, "gpu_allocated_gb": 32.364006912, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 4608.407204627991, "eta_sec": 23.14265300999934, "timestamp": "2026-02-15T15:46:37.999363"} |
|
|
