| {"step": 0, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.8160546875, "mean_total_reward": 0.35518359375, "std_total_reward": 0.5360215104754499, "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": 0.21827483305712755, "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": "2026-02-15T12:50:50.797961"} |
| {"step": 1, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.22, "mean_correctness_reward": 0.22, "mean_length_penalty": -0.94798828125, "mean_total_reward": -0.064396484375, "std_total_reward": 0.43467540176854835, "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": 0.2739858143476051, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 27.934109210968018, "avg_step_time_sec": 27.934109210968018, "gpu_allocated_gb": 32.365621248, "gpu_reserved_gb": 79.92246272, "gpu_peak_gb": 73.906566656, "elapsed_sec": 27.938912868499756, "eta_sec": 5558.8877329826355, "timestamp": "2026-02-15T12:51:20.353387"} |
| {"step": 2, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.7, "mean_correctness_reward": 0.7, "mean_length_penalty": -0.787265625, "mean_total_reward": 0.4638203125, "std_total_reward": 0.5018127322931069, "total_tokens": 80616, "mean_completion_length": 1612.32, "median_completion_length": 1627, "min_completion_length": 745, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.4728567695783963, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 29.54430913925171, "avg_step_time_sec": 28.739209175109863, "gpu_allocated_gb": 32.365654016, "gpu_reserved_gb": 79.95392, "gpu_peak_gb": 74.127090176, "elapsed_sec": 57.49077534675598, "eta_sec": 5690.363416671753, "timestamp": "2026-02-15T12:51:48.007708"} |
| {"step": 3, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.72, "mean_correctness_reward": 0.72, "mean_length_penalty": -0.727255859375, "mean_total_reward": 0.5018232421875, "std_total_reward": 0.5025495390514081, "total_tokens": 74471, "mean_completion_length": 1489.42, "median_completion_length": 1530, "min_completion_length": 626, "max_completion_length": 2048, "frac_has_think": 0.72, "frac_has_boxed": 0.72, "mean_reward_std_per_prompt": 0.2236571805581639, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 27.55850577354431, "avg_step_time_sec": 28.345641374588013, "gpu_allocated_gb": 32.365626368, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.127090176, "elapsed_sec": 85.05627918243408, "eta_sec": 5584.0913507938385, "timestamp": "2026-02-15T12:52:15.562360"} |
| {"step": 4, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.18, "mean_correctness_reward": 0.18, "mean_length_penalty": -0.954775390625, "mean_total_reward": -0.1064326171875, "std_total_reward": 0.4134220202638309, "total_tokens": 97769, "mean_completion_length": 1955.38, "median_completion_length": 2048, "min_completion_length": 1097, "max_completion_length": 2048, "frac_has_think": 0.22, "frac_has_boxed": 0.22, "mean_reward_std_per_prompt": 0.15082109444822508, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 27.57804584503174, "avg_step_time_sec": 28.153742492198944, "gpu_allocated_gb": 32.365638656, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.127090176, "elapsed_sec": 112.64123177528381, "eta_sec": 5518.133528470993, "timestamp": "2026-02-15T12:52:44.709104"} |
| {"step": 5, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.58, "mean_correctness_reward": 0.58, "mean_length_penalty": -0.810546875, "mean_total_reward": 0.3368359375, "std_total_reward": 0.5438899469165082, "total_tokens": 83000, "mean_completion_length": 1660, "median_completion_length": 1925, "min_completion_length": 723, "max_completion_length": 2048, "frac_has_think": 0.68, "frac_has_boxed": 0.68, "mean_reward_std_per_prompt": 0.27244439678553095, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 29.668588161468506, "avg_step_time_sec": 28.456711626052858, "gpu_allocated_gb": 32.365667328, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.156304384, "elapsed_sec": 142.31761074066162, "eta_sec": 5549.058767080307, "timestamp": "2026-02-15T12:53:13.111543"} |
| {"step": 6, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.72, "mean_correctness_reward": 0.72, "mean_length_penalty": -0.77140625, "mean_total_reward": 0.48857812500000003, "std_total_reward": 0.49330821858907525, "total_tokens": 78992, "mean_completion_length": 1579.84, "median_completion_length": 1523, "min_completion_length": 803, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.26053480162502063, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 27.852775812149048, "avg_step_time_sec": 28.35605565706889, "gpu_allocated_gb": 32.365632512, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.156304384, "elapsed_sec": 170.17641854286194, "eta_sec": 5501.074797471365, "timestamp": "2026-02-15T12:53:40.672911"} |
| {"step": 7, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.87951171875, "mean_total_reward": 0.13614648437500002, "std_total_reward": 0.5350195314318154, "total_tokens": 90062, "mean_completion_length": 1801.24, "median_completion_length": 2048, "min_completion_length": 930, "max_completion_length": 2048, "frac_has_think": 0.44, "frac_has_boxed": 0.44, "mean_reward_std_per_prompt": 0.30112940575579167, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 27.557002782821655, "avg_step_time_sec": 28.241905246462142, "gpu_allocated_gb": 32.36562432, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.156304384, "elapsed_sec": 197.7399559020996, "eta_sec": 5450.687712567194, "timestamp": "2026-02-15T12:54:09.512763"} |
| {"step": 8, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.58, "mean_correctness_reward": 0.58, "mean_length_penalty": -0.7813671875, "mean_total_reward": 0.34558984375, "std_total_reward": 0.526313812612997, "total_tokens": 80012, "mean_completion_length": 1600.24, "median_completion_length": 1692, "min_completion_length": 785, "max_completion_length": 2048, "frac_has_think": 0.78, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.2568065284159381, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.726955890655518, "avg_step_time_sec": 28.302536576986313, "gpu_allocated_gb": 32.365616128, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.156304384, "elapsed_sec": 226.47286176681519, "eta_sec": 5434.087022781372, "timestamp": "2026-02-15T12:54:37.045562"} |
| {"step": 9, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.7, "mean_correctness_reward": 0.7, "mean_length_penalty": -0.724658203125, "mean_total_reward": 0.4826025390625, "std_total_reward": 0.5094331465455438, "total_tokens": 74205, "mean_completion_length": 1484.1, "median_completion_length": 1452, "min_completion_length": 668, "max_completion_length": 2048, "frac_has_think": 0.76, "frac_has_boxed": 0.76, "mean_reward_std_per_prompt": 0.2836445981266601, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 27.636572122573853, "avg_step_time_sec": 28.22854052649604, "gpu_allocated_gb": 32.365641728, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.156304384, "elapsed_sec": 254.12109518051147, "eta_sec": 5391.651240560744, "timestamp": "2026-02-15T12:55:04.916783"} |
| {"step": 10, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.58, "mean_correctness_reward": 0.58, "mean_length_penalty": -0.9056640625, "mean_total_reward": 0.30830078125, "std_total_reward": 0.5183093539310046, "total_tokens": 92740, "mean_completion_length": 1854.8, "median_completion_length": 2034, "min_completion_length": 1178, "max_completion_length": 2048, "frac_has_think": 0.68, "frac_has_boxed": 0.68, "mean_reward_std_per_prompt": 0.4623440720553876, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 28.217277765274048, "avg_step_time_sec": 28.22741425037384, "gpu_allocated_gb": 32.365712384, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 282.3446218967438, "eta_sec": 5363.20870757103, "timestamp": "2026-02-15T12:55:34.023201"} |
| {"step": 11, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.5, "mean_correctness_reward": 0.5, "mean_length_penalty": -0.86544921875, "mean_total_reward": 0.24036523437500001, "std_total_reward": 0.541110527572909, "total_tokens": 88622, "mean_completion_length": 1772.44, "median_completion_length": 2048, "min_completion_length": 1000, "max_completion_length": 2048, "frac_has_think": 0.56, "frac_has_boxed": 0.56, "mean_reward_std_per_prompt": 0.3586007011158601, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.757367849349976, "avg_step_time_sec": 28.27559185028076, "gpu_allocated_gb": 32.36563456, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 311.10862851142883, "eta_sec": 5344.086859703064, "timestamp": "2026-02-15T12:56:02.866211"} |
| {"step": 12, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.44, "mean_correctness_reward": 0.44, "mean_length_penalty": -0.84576171875, "mean_total_reward": 0.18627148437500002, "std_total_reward": 0.5284626317703484, "total_tokens": 86606, "mean_completion_length": 1732.12, "median_completion_length": 1925, "min_completion_length": 840, "max_completion_length": 2048, "frac_has_think": 0.66, "frac_has_boxed": 0.66, "mean_reward_std_per_prompt": 0.22041296338652466, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 28.84028148651123, "avg_step_time_sec": 28.322649319966633, "gpu_allocated_gb": 32.365643776, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 339.95515179634094, "eta_sec": 5324.658072153727, "timestamp": "2026-02-15T12:56:31.151649"} |
| {"step": 13, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.52, "mean_correctness_reward": 0.52, "mean_length_penalty": -0.85587890625, "mean_total_reward": 0.26323632812500003, "std_total_reward": 0.5421951341637942, "total_tokens": 87642, "mean_completion_length": 1752.84, "median_completion_length": 2015, "min_completion_length": 975, "max_completion_length": 2048, "frac_has_think": 0.66, "frac_has_boxed": 0.68, "mean_reward_std_per_prompt": 0.3733660990668449, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.278911590576172, "avg_step_time_sec": 28.31928487924429, "gpu_allocated_gb": 32.365641728, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 368.24042987823486, "eta_sec": 5295.706272418682, "timestamp": "2026-02-15T12:56:59.982041"} |
| {"step": 14, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.54, "mean_correctness_reward": 0.54, "mean_length_penalty": -0.7524609375, "mean_total_reward": 0.31426171875000003, "std_total_reward": 0.5190212739873096, "total_tokens": 77052, "mean_completion_length": 1541.04, "median_completion_length": 1481, "min_completion_length": 961, "max_completion_length": 2048, "frac_has_think": 0.88, "frac_has_boxed": 0.88, "mean_reward_std_per_prompt": 0.16386111663786398, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 28.874590396881104, "avg_step_time_sec": 28.35894955907549, "gpu_allocated_gb": 32.365654016, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 397.1227116584778, "eta_sec": 5274.764617988041, "timestamp": "2026-02-15T12:57:27.420071"} |
| {"step": 15, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.78552734375, "mean_total_reward": 0.364341796875, "std_total_reward": 0.5439687035653124, "total_tokens": 80438, "mean_completion_length": 1608.76, "median_completion_length": 1801, "min_completion_length": 763, "max_completion_length": 2048, "frac_has_think": 0.72, "frac_has_boxed": 0.72, "mean_reward_std_per_prompt": 0.2872903598813162, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 27.9766788482666, "avg_step_time_sec": 28.333464845021567, "gpu_allocated_gb": 32.36562944, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 425.10600781440735, "eta_sec": 5241.69099632899, "timestamp": "2026-02-15T12:57:55.713183"} |
| {"step": 16, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.52, "mean_correctness_reward": 0.52, "mean_length_penalty": -0.83439453125, "mean_total_reward": 0.269681640625, "std_total_reward": 0.5484942058441498, "total_tokens": 85442, "mean_completion_length": 1708.84, "median_completion_length": 2028, "min_completion_length": 868, "max_completion_length": 2048, "frac_has_think": 0.62, "frac_has_boxed": 0.62, "mean_reward_std_per_prompt": 0.16993782697205972, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 27.677348852157593, "avg_step_time_sec": 28.292457595467567, "gpu_allocated_gb": 32.365674496, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 452.7897973060608, "eta_sec": 5205.812197566032, "timestamp": "2026-02-15T12:58:23.806670"} |
| {"step": 17, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.62, "mean_correctness_reward": 0.62, "mean_length_penalty": -0.82091796875, "mean_total_reward": 0.373724609375, "std_total_reward": 0.5191164529208405, "total_tokens": 84062, "mean_completion_length": 1681.24, "median_completion_length": 1695, "min_completion_length": 1013, "max_completion_length": 2048, "frac_has_think": 0.9, "frac_has_boxed": 0.9, "mean_reward_std_per_prompt": 0.26385378114765895, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.07231116294861, "avg_step_time_sec": 28.279507805319394, "gpu_allocated_gb": 32.365631488, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 480.8694701194763, "eta_sec": 5175.149928373449, "timestamp": "2026-02-15T12:58:52.261536"} |
| {"step": 18, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.46, "mean_correctness_reward": 0.46, "mean_length_penalty": -0.855869140625, "mean_total_reward": 0.2032392578125, "std_total_reward": 0.5471480306519683, "total_tokens": 87641, "mean_completion_length": 1752.82, "median_completion_length": 2048, "min_completion_length": 589, "max_completion_length": 2048, "frac_has_think": 0.6, "frac_has_boxed": 0.6, "mean_reward_std_per_prompt": 0.3545182317041946, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.46290922164917, "avg_step_time_sec": 28.28969677289327, "gpu_allocated_gb": 32.36563456, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 509.33796882629395, "eta_sec": 5148.7248126665745, "timestamp": "2026-02-15T12:59:20.696570"} |
| {"step": 19, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.48, "mean_correctness_reward": 0.48, "mean_length_penalty": -0.90060546875, "mean_total_reward": 0.209818359375, "std_total_reward": 0.5311301536319002, "total_tokens": 92222, "mean_completion_length": 1844.44, "median_completion_length": 2048, "min_completion_length": 1120, "max_completion_length": 2048, "frac_has_think": 0.58, "frac_has_boxed": 0.58, "mean_reward_std_per_prompt": 0.3660478624732469, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 29.0153169631958, "avg_step_time_sec": 28.32788730922498, "gpu_allocated_gb": 32.365650944, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 538.3603754043579, "eta_sec": 5127.347602969721, "timestamp": "2026-02-15T12:59:49.878422"} |
| {"step": 20, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.7, "mean_correctness_reward": 0.7, "mean_length_penalty": -0.82498046875, "mean_total_reward": 0.452505859375, "std_total_reward": 0.4941444098831394, "total_tokens": 84478, "mean_completion_length": 1689.56, "median_completion_length": 1811, "min_completion_length": 1011, "max_completion_length": 2048, "frac_has_think": 0.78, "frac_has_boxed": 0.78, "mean_reward_std_per_prompt": 0.23446934221160404, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 28.474041223526, "avg_step_time_sec": 28.335195004940033, "gpu_allocated_gb": 32.365620224, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 566.840663433075, "eta_sec": 5100.335100889206, "timestamp": "2026-02-15T13:00:17.030816"} |
| {"step": 21, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.732939453125, "mean_total_reward": 0.5201181640625, "std_total_reward": 0.48812022769661173, "total_tokens": 75053, "mean_completion_length": 1501.06, "median_completion_length": 1416, "min_completion_length": 687, "max_completion_length": 2048, "frac_has_think": 0.78, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.4258443948114295, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 27.308632135391235, "avg_step_time_sec": 28.28631105877104, "gpu_allocated_gb": 32.365652992, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.261332992, "elapsed_sec": 594.1563687324524, "eta_sec": 5063.249679520016, "timestamp": "2026-02-15T13:00:44.321647"} |
| {"step": 22, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.56, "mean_correctness_reward": 0.56, "mean_length_penalty": -0.808720703125, "mean_total_reward": 0.3173837890625, "std_total_reward": 0.53424669436359, "total_tokens": 82813, "mean_completion_length": 1656.26, "median_completion_length": 1648, "min_completion_length": 875, "max_completion_length": 2048, "frac_has_think": 0.72, "frac_has_boxed": 0.72, "mean_reward_std_per_prompt": 0.3385761436242674, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 27.796810388565063, "avg_step_time_sec": 28.264061028307136, "gpu_allocated_gb": 32.365769728, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.392025088, "elapsed_sec": 621.9595530033112, "eta_sec": 5031.0028630386705, "timestamp": "2026-02-15T13:01:12.270448"} |
| {"step": 23, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.54, "mean_correctness_reward": 0.54, "mean_length_penalty": -0.8101171875, "mean_total_reward": 0.29696484375, "std_total_reward": 0.54531423131105, "total_tokens": 82956, "mean_completion_length": 1659.12, "median_completion_length": 1935, "min_completion_length": 734, "max_completion_length": 2048, "frac_has_think": 0.74, "frac_has_boxed": 0.74, "mean_reward_std_per_prompt": 0.3497195702340591, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 28.29141664505005, "avg_step_time_sec": 28.26525040294813, "gpu_allocated_gb": 32.365832192, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 650.2579252719879, "eta_sec": 5002.949321321819, "timestamp": "2026-02-15T13:01:40.807407"} |
| {"step": 24, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.76, "mean_correctness_reward": 0.76, "mean_length_penalty": -0.711455078125, "mean_total_reward": 0.5465634765625, "std_total_reward": 0.4762865727206479, "total_tokens": 72853, "mean_completion_length": 1457.06, "median_completion_length": 1397, "min_completion_length": 754, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.82, "mean_reward_std_per_prompt": 0.2746960922917271, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 27.532538175582886, "avg_step_time_sec": 28.23472072680791, "gpu_allocated_gb": 32.365621248, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 677.7964761257172, "eta_sec": 4969.310847918192, "timestamp": "2026-02-15T13:02:06.934994"} |
| {"step": 25, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.56, "mean_correctness_reward": 0.56, "mean_length_penalty": -0.65728515625, "mean_total_reward": 0.362814453125, "std_total_reward": 0.4920420672102549, "total_tokens": 67306, "mean_completion_length": 1346.12, "median_completion_length": 1283, "min_completion_length": 729, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.1607709082843052, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 26.157861948013306, "avg_step_time_sec": 28.151646375656128, "gpu_allocated_gb": 32.365627392, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 703.9615151882172, "eta_sec": 4926.538115739822, "timestamp": "2026-02-15T13:02:33.155005"} |
| {"step": 26, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.62, "mean_correctness_reward": 0.62, "mean_length_penalty": -0.741142578125, "mean_total_reward": 0.3976572265625, "std_total_reward": 0.5478039525298217, "total_tokens": 75893, "mean_completion_length": 1517.86, "median_completion_length": 1560, "min_completion_length": 606, "max_completion_length": 2048, "frac_has_think": 0.64, "frac_has_boxed": 0.64, "mean_reward_std_per_prompt": 0.07472572471144383, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 26.308947801589966, "avg_step_time_sec": 28.08077335357666, "gpu_allocated_gb": 32.365635584, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 730.2777352333069, "eta_sec": 4886.054563522339, "timestamp": "2026-02-15T13:02:59.642982"} |
| {"step": 27, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.612421875, "mean_total_reward": 0.5562734375, "std_total_reward": 0.4573448147504928, "total_tokens": 62712, "mean_completion_length": 1254.24, "median_completion_length": 1147, "min_completion_length": 796, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.12097272246629967, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 26.284222841262817, "avg_step_time_sec": 28.014234445713186, "gpu_allocated_gb": 32.3656192, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 756.5673427581787, "eta_sec": 4846.462559108381, "timestamp": "2026-02-15T13:03:24.867029"} |
| {"step": 28, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.66, "mean_correctness_reward": 0.66, "mean_length_penalty": -0.789384765625, "mean_total_reward": 0.4231845703125, "std_total_reward": 0.5226410616678681, "total_tokens": 80833, "mean_completion_length": 1616.66, "median_completion_length": 1874, "min_completion_length": 575, "max_completion_length": 2048, "frac_has_think": 0.72, "frac_has_boxed": 0.72, "mean_reward_std_per_prompt": 0.34916670349820844, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.863898992538452, "avg_step_time_sec": 27.937436750956945, "gpu_allocated_gb": 32.365636608, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 782.4379670619965, "eta_sec": 4805.2391211645945, "timestamp": "2026-02-15T13:03:52.483153"} |
| {"step": 29, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.92, "mean_correctness_reward": 0.92, "mean_length_penalty": -0.600908203125, "mean_total_reward": 0.7397275390625, "std_total_reward": 0.310141373190976, "total_tokens": 61533, "mean_completion_length": 1230.66, "median_completion_length": 1109, "min_completion_length": 761, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.12636057952963198, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 27.374521255493164, "avg_step_time_sec": 27.91802587180302, "gpu_allocated_gb": 32.365682688, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 809.819153547287, "eta_sec": 4773.982424078316, "timestamp": "2026-02-15T13:04:18.025848"} |
| {"step": 30, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.32, "mean_correctness_reward": 0.32, "mean_length_penalty": -0.79912109375, "mean_total_reward": 0.08026367187500001, "std_total_reward": 0.4538365624619133, "total_tokens": 81830, "mean_completion_length": 1636.6, "median_completion_length": 1753, "min_completion_length": 671, "max_completion_length": 2048, "frac_has_think": 0.88, "frac_has_boxed": 0.88, "mean_reward_std_per_prompt": 0.29248720156279345, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 25.113107442855835, "avg_step_time_sec": 27.824528590838113, "gpu_allocated_gb": 32.365612032, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 834.939754486084, "eta_sec": 4730.169860442479, "timestamp": "2026-02-15T13:04:45.553664"} |
| {"step": 31, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.84, "mean_correctness_reward": 0.84, "mean_length_penalty": -0.616484375, "mean_total_reward": 0.6550546875, "std_total_reward": 0.4097839566928802, "total_tokens": 63128, "mean_completion_length": 1262.56, "median_completion_length": 1169, "min_completion_length": 595, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.2683408528194143, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 27.76304793357849, "avg_step_time_sec": 27.822545343829738, "gpu_allocated_gb": 32.365669376, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 862.7091100215912, "eta_sec": 4702.010163107226, "timestamp": "2026-02-15T13:05:11.046308"} |
| {"step": 32, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.691787109375, "mean_total_reward": 0.3924638671875, "std_total_reward": 0.5299317131144925, "total_tokens": 70839, "mean_completion_length": 1416.78, "median_completion_length": 1258, "min_completion_length": 765, "max_completion_length": 2048, "frac_has_think": 0.86, "frac_has_boxed": 0.86, "mean_reward_std_per_prompt": 0.19427803746310587, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.24401068687439, "avg_step_time_sec": 27.741966135799885, "gpu_allocated_gb": 32.365612032, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 887.9611225128174, "eta_sec": 4660.650310814381, "timestamp": "2026-02-15T13:05:37.110018"} |
| {"step": 33, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.68, "mean_correctness_reward": 0.68, "mean_length_penalty": -0.72345703125, "mean_total_reward": 0.462962890625, "std_total_reward": 0.520752924306115, "total_tokens": 74082, "mean_completion_length": 1481.64, "median_completion_length": 1465, "min_completion_length": 812, "max_completion_length": 2048, "frac_has_think": 0.72, "frac_has_boxed": 0.72, "mean_reward_std_per_prompt": 0.2997367481110057, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 26.783495664596558, "avg_step_time_sec": 27.71292157606645, "gpu_allocated_gb": 32.365663232, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 914.7510874271393, "eta_sec": 4628.057903203097, "timestamp": "2026-02-15T13:06:04.039095"} |
| {"step": 34, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.678125, "mean_total_reward": 0.5365625, "std_total_reward": 0.4971143655078342, "total_tokens": 69440, "mean_completion_length": 1388.8, "median_completion_length": 1260, "min_completion_length": 717, "max_completion_length": 2048, "frac_has_think": 0.74, "frac_has_boxed": 0.74, "mean_reward_std_per_prompt": 0.28161293869352866, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 26.63880705833435, "avg_step_time_sec": 27.68132997260374, "gpu_allocated_gb": 32.365696, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 941.3970556259155, "eta_sec": 4595.100775452221, "timestamp": "2026-02-15T13:06:30.246553"} |
| {"step": 35, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.5809765625, "mean_total_reward": 0.62570703125, "std_total_reward": 0.45965559954295415, "total_tokens": 59492, "mean_completion_length": 1189.84, "median_completion_length": 1000, "min_completion_length": 626, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.026705656615107733, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.87775206565857, "avg_step_time_sec": 27.629799175262452, "gpu_allocated_gb": 32.365632512, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 967.2803273200989, "eta_sec": 4558.916863918304, "timestamp": "2026-02-15T13:06:55.849242"} |
| {"step": 36, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.56, "mean_correctness_reward": 0.56, "mean_length_penalty": -0.639384765625, "mean_total_reward": 0.3681845703125, "std_total_reward": 0.5299987603200054, "total_tokens": 65473, "mean_completion_length": 1309.46, "median_completion_length": 1141, "min_completion_length": 674, "max_completion_length": 2048, "frac_has_think": 0.78, "frac_has_boxed": 0.78, "mean_reward_std_per_prompt": 0.10569153553229574, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 25.664913177490234, "avg_step_time_sec": 27.57521900865767, "gpu_allocated_gb": 32.365651968, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 992.9516959190369, "eta_sec": 4522.335917419858, "timestamp": "2026-02-15T13:07:21.548394"} |
| {"step": 37, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.82, "mean_correctness_reward": 0.82, "mean_length_penalty": -0.65107421875, "mean_total_reward": 0.624677734375, "std_total_reward": 0.41677440464759136, "total_tokens": 66670, "mean_completion_length": 1333.4, "median_completion_length": 1219, "min_completion_length": 690, "max_completion_length": 2048, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.2733238312810667, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 26.092108249664307, "avg_step_time_sec": 27.53513493409028, "gpu_allocated_gb": 32.36573184, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1019.0503141880035, "eta_sec": 4488.226994256715, "timestamp": "2026-02-15T13:07:47.976412"} |
| {"step": 38, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.58, "mean_correctness_reward": 0.58, "mean_length_penalty": -0.58958984375, "mean_total_reward": 0.403123046875, "std_total_reward": 0.5643443169145521, "total_tokens": 60374, "mean_completion_length": 1207.48, "median_completion_length": 1006, "min_completion_length": 488, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.07966050084377334, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 26.59885048866272, "avg_step_time_sec": 27.51049586973692, "gpu_allocated_gb": 32.365658112, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1045.6553220748901, "eta_sec": 4456.700330897381, "timestamp": "2026-02-15T13:08:13.960260"} |
| {"step": 39, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.455830078125, "mean_total_reward": 0.8632509765624999, "std_total_reward": 0.03326479493732052, "total_tokens": 46677, "mean_completion_length": 933.54, "median_completion_length": 872, "min_completion_length": 641, "max_completion_length": 1624, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.02425146539192231, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.224746227264404, "avg_step_time_sec": 27.45188690454532, "gpu_allocated_gb": 32.365620224, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1070.8876752853394, "eta_sec": 4419.753791631796, "timestamp": "2026-02-15T13:08:35.339426"} |
| {"step": 40, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.5, "mean_correctness_reward": 0.5, "mean_length_penalty": -0.639208984375, "mean_total_reward": 0.3082373046875, "std_total_reward": 0.5367450816484821, "total_tokens": 65455, "mean_completion_length": 1309.1, "median_completion_length": 1184, "min_completion_length": 592, "max_completion_length": 2048, "frac_has_think": 0.86, "frac_has_boxed": 0.86, "mean_reward_std_per_prompt": 0.1290077637127883, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 19.66288185119629, "avg_step_time_sec": 27.257161778211593, "gpu_allocated_gb": 32.36502784, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1090.557326555252, "eta_sec": 4361.145884513855, "timestamp": "2026-02-15T13:08:59.231762"} |
| {"step": 41, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.84, "mean_correctness_reward": 0.84, "mean_length_penalty": -0.6017578125, "mean_total_reward": 0.65947265625, "std_total_reward": 0.42282164908446834, "total_tokens": 61620, "mean_completion_length": 1232.4, "median_completion_length": 1150, "min_completion_length": 473, "max_completion_length": 2048, "frac_has_think": 0.86, "frac_has_boxed": 0.86, "mean_reward_std_per_prompt": 0.17587924310085562, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 26.634891510009766, "avg_step_time_sec": 27.241984454596913, "gpu_allocated_gb": 32.365696, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1117.198964357376, "eta_sec": 4331.475528280909, "timestamp": "2026-02-15T13:09:25.486670"} |
| {"step": 42, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.92, "mean_correctness_reward": 0.92, "mean_length_penalty": -0.43365234375, "mean_total_reward": 0.789904296875, "std_total_reward": 0.2829412012045627, "total_tokens": 44406, "mean_completion_length": 888.12, "median_completion_length": 748, "min_completion_length": 459, "max_completion_length": 2048, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.16961050898627317, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.381516456604004, "avg_step_time_sec": 27.197687597501847, "gpu_allocated_gb": 32.365651968, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1142.5876851081848, "eta_sec": 4297.234640405291, "timestamp": "2026-02-15T13:09:49.106674"} |
| {"step": 43, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.86, "mean_correctness_reward": 0.86, "mean_length_penalty": -0.57484375, "mean_total_reward": 0.687546875, "std_total_reward": 0.4008690223831024, "total_tokens": 58864, "mean_completion_length": 1177.28, "median_completion_length": 1101, "min_completion_length": 586, "max_completion_length": 2048, "frac_has_think": 0.92, "frac_has_boxed": 0.92, "mean_reward_std_per_prompt": 0.22376753611427452, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 23.389315843582153, "avg_step_time_sec": 27.109120812526967, "gpu_allocated_gb": 32.365605888, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1165.9838280677795, "eta_sec": 4256.131967566734, "timestamp": "2026-02-15T13:10:14.037153"} |
| {"step": 44, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.72, "mean_correctness_reward": 0.72, "mean_length_penalty": -0.563359375, "mean_total_reward": 0.5509921875, "std_total_reward": 0.5140042939730342, "total_tokens": 57688, "mean_completion_length": 1153.76, "median_completion_length": 940, "min_completion_length": 451, "max_completion_length": 2048, "frac_has_think": 0.88, "frac_has_boxed": 0.88, "mean_reward_std_per_prompt": 0.12319801940042935, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.58506679534912, "avg_step_time_sec": 27.07448322122747, "gpu_allocated_gb": 32.365645824, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1191.5760550498962, "eta_sec": 4223.619382511485, "timestamp": "2026-02-15T13:10:39.537970"} |
| {"step": 45, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.5759765625, "mean_total_reward": 0.62720703125, "std_total_reward": 0.46453745027234983, "total_tokens": 58980, "mean_completion_length": 1179.6, "median_completion_length": 1059, "min_completion_length": 551, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.015025911328373994, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.00561499595642, "avg_step_time_sec": 27.028508371777004, "gpu_allocated_gb": 32.365638656, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1216.5881185531616, "eta_sec": 4189.418797625435, "timestamp": "2026-02-15T13:11:04.672436"} |
| {"step": 46, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.9, "mean_correctness_reward": 0.9, "mean_length_penalty": -0.460947265625, "mean_total_reward": 0.7617158203125, "std_total_reward": 0.33643380532458533, "total_tokens": 47201, "mean_completion_length": 944.02, "median_completion_length": 790, "min_completion_length": 576, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.11061277653975822, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 25.208638429641724, "avg_step_time_sec": 26.988945981730584, "gpu_allocated_gb": 32.3656704, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1241.803980588913, "eta_sec": 4156.29768118651, "timestamp": "2026-02-15T13:11:28.745002"} |
| {"step": 47, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.68, "mean_correctness_reward": 0.68, "mean_length_penalty": -0.556943359375, "mean_total_reward": 0.5129169921875, "std_total_reward": 0.5012127268538882, "total_tokens": 57031, "mean_completion_length": 1140.62, "median_completion_length": 1050, "min_completion_length": 514, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.1966050021222829, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.492669820785522, "avg_step_time_sec": 26.935833722987073, "gpu_allocated_gb": 32.36563456, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1266.3033528327942, "eta_sec": 4121.182559617022, "timestamp": "2026-02-15T13:11:54.130632"} |
| {"step": 48, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.532412109375, "mean_total_reward": 0.6402763671875, "std_total_reward": 0.4709574636833646, "total_tokens": 54519, "mean_completion_length": 1090.38, "median_completion_length": 959, "min_completion_length": 518, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.014771836823322549, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.876787900924683, "avg_step_time_sec": 26.89293693502744, "gpu_allocated_gb": 32.3656448, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1291.1865510940552, "eta_sec": 4087.726414124171, "timestamp": "2026-02-15T13:12:18.786768"} |
| {"step": 49, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.479462890625, "mean_total_reward": 0.4561611328125, "std_total_reward": 0.5509277749409011, "total_tokens": 49097, "mean_completion_length": 981.94, "median_completion_length": 834, "min_completion_length": 329, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.009023401770548478, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 24.644437789916992, "avg_step_time_sec": 26.847049197372126, "gpu_allocated_gb": 32.365645824, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1315.8366305828094, "eta_sec": 4053.904428803191, "timestamp": "2026-02-15T13:12:43.068581"} |
| {"step": 50, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.4571875, "mean_total_reward": 0.86284375, "std_total_reward": 0.05283686687233649, "total_tokens": 46816, "mean_completion_length": 936.32, "median_completion_length": 808, "min_completion_length": 428, "max_completion_length": 1852, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.02474990931734199, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.730982542037964, "avg_step_time_sec": 26.80472786426544, "gpu_allocated_gb": 32.365627392, "gpu_reserved_gb": 79.956017152, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1340.5745964050293, "eta_sec": 4020.7091796398163, "timestamp": "2026-02-15T13:13:07.036644"} |
| {"step": 51, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.409609375, "mean_total_reward": 0.6771171875, "std_total_reward": 0.39906075279432873, "total_tokens": 41944, "mean_completion_length": 838.88, "median_completion_length": 799, "min_completion_length": 474, "max_completion_length": 1391, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.010492119691384097, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 21.89322590827942, "avg_step_time_sec": 26.708423904344148, "gpu_allocated_gb": 32.36534784, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1363.1216430664062, "eta_sec": 3979.555161747278, "timestamp": "2026-02-15T13:13:25.744559"} |
| {"step": 52, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.7, "mean_correctness_reward": 0.7, "mean_length_penalty": -0.504482421875, "mean_total_reward": 0.5486552734375, "std_total_reward": 0.5022524073027077, "total_tokens": 51659, "mean_completion_length": 1033.18, "median_completion_length": 952, "min_completion_length": 453, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.1721635802689769, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 16.877690315246582, "avg_step_time_sec": 26.51937133532304, "gpu_allocated_gb": 32.36470272, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1380.0069017410278, "eta_sec": 3924.86695762781, "timestamp": "2026-02-15T13:13:47.435643"} |
| {"step": 53, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.76, "mean_correctness_reward": 0.76, "mean_length_penalty": -0.61236328125, "mean_total_reward": 0.576291015625, "std_total_reward": 0.4793199990720891, "total_tokens": 62706, "mean_completion_length": 1254.12, "median_completion_length": 1210, "min_completion_length": 588, "max_completion_length": 2048, "frac_has_think": 0.82, "frac_has_boxed": 0.82, "mean_reward_std_per_prompt": 0.22294172562710424, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 24.47743272781372, "avg_step_time_sec": 26.480844191785128, "gpu_allocated_gb": 32.365645824, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1404.4919390678406, "eta_sec": 3892.6840961924136, "timestamp": "2026-02-15T13:14:13.603833"} |
| {"step": 54, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.7, "mean_correctness_reward": 0.7, "mean_length_penalty": -0.6122265625, "mean_total_reward": 0.51633203125, "std_total_reward": 0.5353445826013994, "total_tokens": 62692, "mean_completion_length": 1253.84, "median_completion_length": 982, "min_completion_length": 545, "max_completion_length": 2048, "frac_has_think": 0.76, "frac_has_boxed": 0.76, "mean_reward_std_per_prompt": 0.1839631301616125, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 26.22730326652527, "avg_step_time_sec": 26.4761489894655, "gpu_allocated_gb": 32.365657088, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1430.7259948253632, "eta_sec": 3865.517752461963, "timestamp": "2026-02-15T13:14:39.310505"} |
| {"step": 55, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.48220703125, "mean_total_reward": 0.455337890625, "std_total_reward": 0.5326768284621313, "total_tokens": 49378, "mean_completion_length": 987.56, "median_completion_length": 671, "min_completion_length": 398, "max_completion_length": 2048, "frac_has_think": 0.82, "frac_has_boxed": 0.82, "mean_reward_std_per_prompt": 0.008148210657675375, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 25.63522434234619, "avg_step_time_sec": 26.46085945042697, "gpu_allocated_gb": 32.365641728, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1456.3684978485107, "eta_sec": 3836.8246203119106, "timestamp": "2026-02-15T13:15:03.850770"} |
| {"step": 56, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.577841796875, "mean_total_reward": 0.6266474609375, "std_total_reward": 0.4676845947821794, "total_tokens": 59171, "mean_completion_length": 1183.42, "median_completion_length": 836, "min_completion_length": 448, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.01316942682494476, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.532681226730347, "avg_step_time_sec": 26.426427696432388, "gpu_allocated_gb": 32.365648896, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1480.9074413776398, "eta_sec": 3805.405588286264, "timestamp": "2026-02-15T13:15:29.742942"} |
| {"step": 57, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.64, "mean_correctness_reward": 0.64, "mean_length_penalty": -0.641953125, "mean_total_reward": 0.4474140625, "std_total_reward": 0.5512727458491028, "total_tokens": 65736, "mean_completion_length": 1314.72, "median_completion_length": 1406, "min_completion_length": 486, "max_completion_length": 2048, "frac_has_think": 0.78, "frac_has_boxed": 0.78, "mean_reward_std_per_prompt": 0.17709471659473955, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 25.889875888824463, "avg_step_time_sec": 26.41701450682523, "gpu_allocated_gb": 32.365647872, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1506.8046097755432, "eta_sec": 3777.6330744760076, "timestamp": "2026-02-15T13:15:55.694986"} |
| {"step": 58, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.88, "mean_correctness_reward": 0.88, "mean_length_penalty": -0.44392578125, "mean_total_reward": 0.746822265625, "std_total_reward": 0.37063457313769604, "total_tokens": 45458, "mean_completion_length": 909.16, "median_completion_length": 833, "min_completion_length": 438, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.11667792730062494, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 26.399674654006958, "avg_step_time_sec": 26.416715543845605, "gpu_allocated_gb": 32.36573696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1533.2114446163177, "eta_sec": 3751.173607226076, "timestamp": "2026-02-15T13:16:20.177317"} |
| {"step": 59, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.38, "mean_correctness_reward": 0.38, "mean_length_penalty": -0.436494140625, "mean_total_reward": 0.2490517578125, "std_total_reward": 0.48914862131901066, "total_tokens": 44697, "mean_completion_length": 893.94, "median_completion_length": 930, "min_completion_length": 389, "max_completion_length": 1467, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.07295113743622943, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 24.44585871696472, "avg_step_time_sec": 26.383311190847625, "gpu_allocated_gb": 32.365630464, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1557.664921760559, "eta_sec": 3720.046877909515, "timestamp": "2026-02-15T13:16:41.014884"} |
| {"step": 60, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.3988671875, "mean_total_reward": 0.88033984375, "std_total_reward": 0.02427402368187045, "total_tokens": 40844, "mean_completion_length": 816.88, "median_completion_length": 810, "min_completion_length": 530, "max_completion_length": 1332, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.015036356972217557, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 18.145666122436523, "avg_step_time_sec": 26.246017106374104, "gpu_allocated_gb": 32.364839936, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1575.8170683383942, "eta_sec": 3674.4423948923745, "timestamp": "2026-02-15T13:16:58.028534"} |
| {"step": 61, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.84, "mean_correctness_reward": 0.84, "mean_length_penalty": -0.508076171875, "mean_total_reward": 0.6875771484375, "std_total_reward": 0.40851658903567384, "total_tokens": 52027, "mean_completion_length": 1040.54, "median_completion_length": 1041, "min_completion_length": 389, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.2595542175437528, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 16.19462490081787, "avg_step_time_sec": 26.081240184971545, "gpu_allocated_gb": 32.364619264, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1592.0187628269196, "eta_sec": 3625.292385711045, "timestamp": "2026-02-15T13:17:19.438830"} |
| {"step": 62, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.404677734375, "mean_total_reward": 0.6785966796875, "std_total_reward": 0.44246242758189414, "total_tokens": 41439, "mean_completion_length": 828.78, "median_completion_length": 818, "min_completion_length": 316, "max_completion_length": 1687, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.013563783000341875, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.528756618499756, "avg_step_time_sec": 26.056200127447806, "gpu_allocated_gb": 32.365659136, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1616.5553467273712, "eta_sec": 3595.755617587797, "timestamp": "2026-02-15T13:17:41.424362"} |
| {"step": 63, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.94, "mean_correctness_reward": 0.94, "mean_length_penalty": -0.430439453125, "mean_total_reward": 0.8108681640625, "std_total_reward": 0.24358679136231012, "total_tokens": 44077, "mean_completion_length": 881.54, "median_completion_length": 858, "min_completion_length": 503, "max_completion_length": 2002, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.14416314117302748, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 20.423675298690796, "avg_step_time_sec": 25.96679497143579, "gpu_allocated_gb": 32.365147648, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1636.9858937263489, "eta_sec": 3557.450911086703, "timestamp": "2026-02-15T13:18:03.069835"} |
| {"step": 64, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.82, "mean_correctness_reward": 0.82, "mean_length_penalty": -0.488037109375, "mean_total_reward": 0.6735888671875, "std_total_reward": 0.44327206346299586, "total_tokens": 49975, "mean_completion_length": 999.5, "median_completion_length": 885, "min_completion_length": 454, "max_completion_length": 2048, "frac_has_think": 0.88, "frac_has_boxed": 0.88, "mean_reward_std_per_prompt": 0.15916536750447674, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 22.932844877243042, "avg_step_time_sec": 25.919389501214027, "gpu_allocated_gb": 32.365581824, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1659.9268193244934, "eta_sec": 3525.0369721651077, "timestamp": "2026-02-15T13:18:27.223735"} |
| {"step": 65, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.4274609375, "mean_total_reward": 0.67176171875, "std_total_reward": 0.4770779793335894, "total_tokens": 43772, "mean_completion_length": 875.44, "median_completion_length": 683, "min_completion_length": 362, "max_completion_length": 2048, "frac_has_think": 0.88, "frac_has_boxed": 0.88, "mean_reward_std_per_prompt": 0.016429975905563148, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 24.40627145767212, "avg_step_time_sec": 25.896110762082614, "gpu_allocated_gb": 32.36566016, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1684.3408296108246, "eta_sec": 3495.974952881153, "timestamp": "2026-02-15T13:18:51.786175"} |
| {"step": 66, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.404072265625, "mean_total_reward": 0.4787783203125, "std_total_reward": 0.5037492927082482, "total_tokens": 41377, "mean_completion_length": 827.54, "median_completion_length": 881, "min_completion_length": 275, "max_completion_length": 1456, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.013141785196109286, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 24.537346601486206, "avg_step_time_sec": 25.875523426316, "gpu_allocated_gb": 32.36565504, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1708.8850758075714, "eta_sec": 3467.320139126344, "timestamp": "2026-02-15T13:19:11.742608"} |
| {"step": 67, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.42572265625, "mean_total_reward": 0.672283203125, "std_total_reward": 0.43424413427806624, "total_tokens": 43594, "mean_completion_length": 871.88, "median_completion_length": 794, "min_completion_length": 508, "max_completion_length": 1989, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0176409358794995, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 17.337170600891113, "avg_step_time_sec": 25.748085324443988, "gpu_allocated_gb": 32.364808704, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1726.2296087741852, "eta_sec": 3424.49534815105, "timestamp": "2026-02-15T13:19:32.275734"} |
| {"step": 68, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.416259765625, "mean_total_reward": 0.47512207031250003, "std_total_reward": 0.5434139475399283, "total_tokens": 42625, "mean_completion_length": 852.5, "median_completion_length": 574, "min_completion_length": 426, "max_completion_length": 2048, "frac_has_think": 0.84, "frac_has_boxed": 0.84, "mean_reward_std_per_prompt": 0.014006964742084732, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 22.87362027168274, "avg_step_time_sec": 25.70581377955044, "gpu_allocated_gb": 32.36556544, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1749.109831571579, "eta_sec": 3393.167418900658, "timestamp": "2026-02-15T13:19:55.861139"} |
| {"step": 69, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.2, "mean_correctness_reward": 0.2, "mean_length_penalty": -0.528798828125, "mean_total_reward": 0.04136035156250001, "std_total_reward": 0.4403538623071282, "total_tokens": 54149, "mean_completion_length": 1082.98, "median_completion_length": 781, "min_completion_length": 435, "max_completion_length": 2048, "frac_has_think": 0.8, "frac_has_boxed": 0.8, "mean_reward_std_per_prompt": 0.011913295735249598, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 23.83001208305359, "avg_step_time_sec": 25.678628247717153, "gpu_allocated_gb": 32.365641728, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1772.946753025055, "eta_sec": 3363.9003004509473, "timestamp": "2026-02-15T13:20:21.187348"} |
| {"step": 70, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.54, "mean_correctness_reward": 0.54, "mean_length_penalty": -0.331025390625, "mean_total_reward": 0.4406923828125, "std_total_reward": 0.514069803100469, "total_tokens": 33897, "mean_completion_length": 677.94, "median_completion_length": 630, "min_completion_length": 333, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.11184319196909695, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 25.331684589385986, "avg_step_time_sec": 25.673671909740992, "gpu_allocated_gb": 32.36564992, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1798.284775018692, "eta_sec": 3337.577348266329, "timestamp": "2026-02-15T13:20:44.060537"} |
| {"step": 71, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.311796875, "mean_total_reward": 0.6464609375, "std_total_reward": 0.45587489191654934, "total_tokens": 31928, "mean_completion_length": 638.56, "median_completion_length": 577, "min_completion_length": 352, "max_completion_length": 1150, "frac_has_think": 0.98, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.09994044764093303, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 22.678415060043335, "avg_step_time_sec": 25.63148519354807, "gpu_allocated_gb": 32.365621248, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1820.9703030586243, "eta_sec": 3306.461589967701, "timestamp": "2026-02-15T13:21:01.337858"} |
| {"step": 72, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.66, "mean_correctness_reward": 0.66, "mean_length_penalty": -0.43396484375, "mean_total_reward": 0.529810546875, "std_total_reward": 0.5152516176512089, "total_tokens": 44438, "mean_completion_length": 888.76, "median_completion_length": 645, "min_completion_length": 440, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.10137096392082953, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 13.84504508972168, "avg_step_time_sec": 25.46778463655048, "gpu_allocated_gb": 32.364408832, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1834.8215990066528, "eta_sec": 3259.8764334784614, "timestamp": "2026-02-15T13:21:21.622768"} |
| {"step": 73, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.31302734375, "mean_total_reward": 0.306091796875, "std_total_reward": 0.5041699759323762, "total_tokens": 32054, "mean_completion_length": 641.08, "median_completion_length": 583, "min_completion_length": 316, "max_completion_length": 1745, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.017801409130660164, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 24.37180995941162, "avg_step_time_sec": 25.45277128480885, "gpu_allocated_gb": 32.36562944, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1859.199390888214, "eta_sec": 3232.501953170724, "timestamp": "2026-02-15T13:21:42.825565"} |
| {"step": 74, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.30919921875, "mean_total_reward": 0.707240234375, "std_total_reward": 0.4104285410389328, "total_tokens": 31662, "mean_completion_length": 633.24, "median_completion_length": 550, "min_completion_length": 227, "max_completion_length": 1529, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.12967954259798625, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 19.447922706604004, "avg_step_time_sec": 25.37162468240068, "gpu_allocated_gb": 32.3652352, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1878.654578924179, "eta_sec": 3196.8247099824857, "timestamp": "2026-02-15T13:22:01.164442"} |
| {"step": 75, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.76, "mean_correctness_reward": 0.76, "mean_length_penalty": -0.321318359375, "mean_total_reward": 0.6636044921875, "std_total_reward": 0.4853002107605633, "total_tokens": 32903, "mean_completion_length": 658.06, "median_completion_length": 527, "min_completion_length": 172, "max_completion_length": 2048, "frac_has_think": 0.92, "frac_has_boxed": 0.92, "mean_reward_std_per_prompt": 0.137181337190251, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 17.334580659866333, "avg_step_time_sec": 25.264464095433553, "gpu_allocated_gb": 32.364903424, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1895.995369195938, "eta_sec": 3158.058011929194, "timestamp": "2026-02-15T13:22:21.981154"} |
| {"step": 76, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.86, "mean_correctness_reward": 0.86, "mean_length_penalty": -0.307021484375, "mean_total_reward": 0.7678935546875, "std_total_reward": 0.4022891133320065, "total_tokens": 31439, "mean_completion_length": 628.78, "median_completion_length": 455, "min_completion_length": 214, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.10152133248493768, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 23.092844247817993, "avg_step_time_sec": 25.23589015007019, "gpu_allocated_gb": 32.365657088, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1919.0949747562408, "eta_sec": 3129.2503786087036, "timestamp": "2026-02-15T13:22:44.964812"} |
| {"step": 77, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.56, "mean_correctness_reward": 0.56, "mean_length_penalty": -0.292802734375, "mean_total_reward": 0.4721591796875, "std_total_reward": 0.5181223153195467, "total_tokens": 29983, "mean_completion_length": 599.66, "median_completion_length": 552, "min_completion_length": 300, "max_completion_length": 1658, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.09488927764873531, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 22.860948085784912, "avg_step_time_sec": 25.205046746637915, "gpu_allocated_gb": 32.36562432, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1941.9626746177673, "eta_sec": 3100.2207498364637, "timestamp": "2026-02-15T13:23:05.626039"} |
| {"step": 78, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.76, "mean_correctness_reward": 0.76, "mean_length_penalty": -0.317001953125, "mean_total_reward": 0.6648994140625, "std_total_reward": 0.4261697352407918, "total_tokens": 32461, "mean_completion_length": 649.22, "median_completion_length": 661, "min_completion_length": 359, "max_completion_length": 1229, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.12722797845683353, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 19.008667945861816, "avg_step_time_sec": 25.12560599278181, "gpu_allocated_gb": 32.365080576, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1960.9779102802277, "eta_sec": 3065.323931119381, "timestamp": "2026-02-15T13:23:21.703357"} |
| {"step": 79, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.294638671875, "mean_total_reward": 0.3116083984375, "std_total_reward": 0.5051528274330515, "total_tokens": 30171, "mean_completion_length": 603.42, "median_completion_length": 616, "min_completion_length": 246, "max_completion_length": 1211, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.012889433464656716, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 14.47655463218689, "avg_step_time_sec": 24.990807874293267, "gpu_allocated_gb": 32.364503552, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1975.4603440761566, "eta_sec": 3023.887752789485, "timestamp": "2026-02-15T13:23:35.840941"} |
| {"step": 80, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.317099609375, "mean_total_reward": 0.6448701171875, "std_total_reward": 0.41928506076204325, "total_tokens": 32471, "mean_completion_length": 649.42, "median_completion_length": 720, "min_completion_length": 292, "max_completion_length": 1088, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.1473599203367662, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 14.095787048339844, "avg_step_time_sec": 24.85462011396885, "gpu_allocated_gb": 32.364481536, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 1989.5631334781647, "eta_sec": 2982.554413676262, "timestamp": "2026-02-15T13:23:49.661756"} |
| {"step": 81, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.217080078125, "mean_total_reward": 0.7348759765625, "std_total_reward": 0.41744264549171733, "total_tokens": 22229, "mean_completion_length": 444.58, "median_completion_length": 415, "min_completion_length": 238, "max_completion_length": 748, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.006412728458098613, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 13.667572975158691, "avg_step_time_sec": 24.716508420897117, "gpu_allocated_gb": 32.364391936, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2003.237817287445, "eta_sec": 2941.264502086757, "timestamp": "2026-02-15T13:24:00.972061"} |
| {"step": 82, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.56, "mean_correctness_reward": 0.56, "mean_length_penalty": -0.1992578125, "mean_total_reward": 0.50022265625, "std_total_reward": 0.49447081862574266, "total_tokens": 20404, "mean_completion_length": 408.08, "median_completion_length": 383, "min_completion_length": 237, "max_completion_length": 720, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.12443847519417361, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 9.647696733474731, "avg_step_time_sec": 24.532742424709042, "gpu_allocated_gb": 32.363818496, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2012.891767501831, "eta_sec": 2894.863606115667, "timestamp": "2026-02-15T13:24:09.754971"} |
| {"step": 83, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.25919921875, "mean_total_reward": 0.722240234375, "std_total_reward": 0.3880948263826563, "total_tokens": 26542, "mean_completion_length": 530.84, "median_completion_length": 361, "min_completion_length": 193, "max_completion_length": 1560, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.008675094940944275, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 8.677106380462646, "avg_step_time_sec": 24.341710665139807, "gpu_allocated_gb": 32.363772928, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2021.5753798484802, "eta_sec": 2847.9801478213576, "timestamp": "2026-02-15T13:24:23.894508"} |
| {"step": 84, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.68, "mean_correctness_reward": 0.68, "mean_length_penalty": -0.34220703125, "mean_total_reward": 0.577337890625, "std_total_reward": 0.48404226366844555, "total_tokens": 35042, "mean_completion_length": 700.84, "median_completion_length": 419, "min_completion_length": 203, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.1950791776105818, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 17.43160343170166, "avg_step_time_sec": 24.25944748378935, "gpu_allocated_gb": 32.364978176, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2039.0150020122528, "eta_sec": 2814.0959081195647, "timestamp": "2026-02-15T13:24:45.097744"} |
| {"step": 85, "num_prompts": 5, "num_rollouts": 50, "accuracy": 1.0, "mean_correctness_reward": 1.0, "mean_length_penalty": -0.151689453125, "mean_total_reward": 0.9544931640625, "std_total_reward": 0.013940806762085022, "total_tokens": 15533, "mean_completion_length": 310.66, "median_completion_length": 302, "min_completion_length": 151, "max_completion_length": 527, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.004140746775359694, "frac_prompts_all_correct": 1.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 22.951135396957397, "avg_step_time_sec": 24.244055576885447, "gpu_allocated_gb": 32.365607936, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2061.9721207618713, "eta_sec": 2788.0663913418266, "timestamp": "2026-02-15T13:24:57.359022"} |
| {"step": 86, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.243369140625, "mean_total_reward": 0.5269892578125001, "std_total_reward": 0.5070043640823568, "total_tokens": 24921, "mean_completion_length": 498.42, "median_completion_length": 420, "min_completion_length": 146, "max_completion_length": 1127, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.009162579205471139, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 6.470686197280884, "avg_step_time_sec": 24.037388491076094, "gpu_allocated_gb": 32.36348416, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2068.449390888214, "eta_sec": 2740.2622879826745, "timestamp": "2026-02-15T13:25:08.015462"} |
| {"step": 87, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.96, "mean_correctness_reward": 0.96, "mean_length_penalty": -0.1865625, "mean_total_reward": 0.90403125, "std_total_reward": 0.1993573217799736, "total_tokens": 19104, "mean_completion_length": 382.08, "median_completion_length": 396, "min_completion_length": 136, "max_completion_length": 679, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.12397853877578585, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 13.484757661819458, "avg_step_time_sec": 23.916093883843256, "gpu_allocated_gb": 32.36433152, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2081.944858789444, "eta_sec": 2702.518608874288, "timestamp": "2026-02-15T13:25:18.616607"} |
| {"step": 88, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.52, "mean_correctness_reward": 0.52, "mean_length_penalty": -0.302001953125, "mean_total_reward": 0.4293994140625, "std_total_reward": 0.542515200498661, "total_tokens": 30925, "mean_completion_length": 618.5, "median_completion_length": 539, "min_completion_length": 197, "max_completion_length": 2048, "frac_has_think": 0.98, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.21429300364494197, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 8.45608139038086, "avg_step_time_sec": 23.740411923690274, "gpu_allocated_gb": 32.363739136, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2090.4073989391327, "eta_sec": 2658.926135453311, "timestamp": "2026-02-15T13:25:36.285158"} |
| {"step": 89, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.94, "mean_correctness_reward": 0.94, "mean_length_penalty": -0.322138671875, "mean_total_reward": 0.8433583984375, "std_total_reward": 0.26080965859725413, "total_tokens": 32987, "mean_completion_length": 659.74, "median_completion_length": 413, "min_completion_length": 274, "max_completion_length": 1998, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.18783953024654348, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 23.021668672561646, "avg_step_time_sec": 23.732336156823663, "gpu_allocated_gb": 32.365671424, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2113.437860250473, "eta_sec": 2634.2893134074266, "timestamp": "2026-02-15T13:25:59.045329"} |
| {"step": 90, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.94, "mean_correctness_reward": 0.94, "mean_length_penalty": -0.25248046875, "mean_total_reward": 0.864255859375, "std_total_reward": 0.26625781320202696, "total_tokens": 25854, "mean_completion_length": 517.08, "median_completion_length": 331, "min_completion_length": 89, "max_completion_length": 1604, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.14807373574870172, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 22.452823638916016, "avg_step_time_sec": 23.71811935106913, "gpu_allocated_gb": 32.365568512, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2135.8977904319763, "eta_sec": 2608.9931286176043, "timestamp": "2026-02-15T13:26:18.936049"} |
| {"step": 91, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.249619140625, "mean_total_reward": 0.7251142578125, "std_total_reward": 0.3936012721791778, "total_tokens": 25561, "mean_completion_length": 511.22, "median_completion_length": 364, "min_completion_length": 156, "max_completion_length": 2048, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.01287620114780105, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 18.874451160430908, "avg_step_time_sec": 23.664892228095088, "gpu_allocated_gb": 32.3651712, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2154.779331922531, "eta_sec": 2579.4732528623645, "timestamp": "2026-02-15T13:26:40.707765"} |
| {"step": 92, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.326708984375, "mean_total_reward": 0.7019873046875, "std_total_reward": 0.44489036583791586, "total_tokens": 33455, "mean_completion_length": 669.1, "median_completion_length": 534, "min_completion_length": 187, "max_completion_length": 2048, "frac_has_think": 0.94, "frac_has_boxed": 0.94, "mean_reward_std_per_prompt": 0.030960502550944433, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 23.017970085144043, "avg_step_time_sec": 23.657860465671707, "gpu_allocated_gb": 32.365626368, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2177.80414891243, "eta_sec": 2555.0489302925444, "timestamp": "2026-02-15T13:27:04.284156"} |
| {"step": 93, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.52, "mean_correctness_reward": 0.52, "mean_length_penalty": -0.21984375, "mean_total_reward": 0.454046875, "std_total_reward": 0.48495847756704064, "total_tokens": 22512, "mean_completion_length": 450.24, "median_completion_length": 340, "min_completion_length": 215, "max_completion_length": 1176, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.16461967861813057, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 24.103271007537842, "avg_step_time_sec": 23.66264982633693, "gpu_allocated_gb": 32.365775872, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2201.913476228714, "eta_sec": 2531.9035314180514, "timestamp": "2026-02-15T13:27:21.923681"} |
| {"step": 94, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.141640625, "mean_total_reward": 0.9375078125, "std_total_reward": 0.13976033041890418, "total_tokens": 14504, "mean_completion_length": 290.08, "median_completion_length": 289, "min_completion_length": 155, "max_completion_length": 459, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06440135783692029, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 14.290348768234253, "avg_step_time_sec": 23.562944495931585, "gpu_allocated_gb": 32.364602368, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2216.210077524185, "eta_sec": 2497.672116568748, "timestamp": "2026-02-15T13:27:31.166843"} |
| {"step": 95, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.68, "mean_correctness_reward": 0.68, "mean_length_penalty": -0.152734375, "mean_total_reward": 0.6341796875, "std_total_reward": 0.47763361354669426, "total_tokens": 15640, "mean_completion_length": 312.8, "median_completion_length": 282, "min_completion_length": 101, "max_completion_length": 684, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.2368899793136358, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 5.872462272644043, "avg_step_time_sec": 23.37672889358119, "gpu_allocated_gb": 32.363398144, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2222.0887413024902, "eta_sec": 2454.556533826025, "timestamp": "2026-02-15T13:27:38.484895"} |
| {"step": 96, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.1479296875, "mean_total_reward": 0.93562109375, "std_total_reward": 0.13823036671136502, "total_tokens": 15148, "mean_completion_length": 302.96, "median_completion_length": 283, "min_completion_length": 133, "max_completion_length": 546, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06369091835108054, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 8.25272798538208, "avg_step_time_sec": 23.219187217454117, "gpu_allocated_gb": 32.36374528, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2230.3474156856537, "eta_sec": 2414.795470615228, "timestamp": "2026-02-15T13:27:45.914202"} |
| {"step": 97, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.17078125, "mean_total_reward": 0.348765625, "std_total_reward": 0.521084526109858, "total_tokens": 17488, "mean_completion_length": 349.76, "median_completion_length": 150, "min_completion_length": 74, "max_completion_length": 2048, "frac_has_think": 0.96, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.02191173006187798, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 7.368400573730469, "avg_step_time_sec": 23.055777045869334, "gpu_allocated_gb": 32.363542528, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2237.7224037647247, "eta_sec": 2374.7450357245416, "timestamp": "2026-02-15T13:28:02.513042"} |
| {"step": 98, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.72, "mean_correctness_reward": 0.72, "mean_length_penalty": -0.090751953125, "mean_total_reward": 0.6927744140625, "std_total_reward": 0.45109511504275607, "total_tokens": 9293, "mean_completion_length": 185.86, "median_completion_length": 178, "min_completion_length": 92, "max_completion_length": 310, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.10057291632836914, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 21.843141078948975, "avg_step_time_sec": 23.04340320947219, "gpu_allocated_gb": 32.365626368, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2259.5717554092407, "eta_sec": 2350.4271273661634, "timestamp": "2026-02-15T13:28:13.284668"} |
| {"step": 99, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.98, "mean_correctness_reward": 0.98, "mean_length_penalty": -0.091435546875, "mean_total_reward": 0.9525693359375, "std_total_reward": 0.13988550882348041, "total_tokens": 9363, "mean_completion_length": 187.26, "median_completion_length": 169, "min_completion_length": 77, "max_completion_length": 346, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06441284534466774, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 4.203660249710083, "avg_step_time_sec": 22.8531027755352, "gpu_allocated_gb": 32.363210752, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2263.781728744507, "eta_sec": 2308.163380329055, "timestamp": "2026-02-15T13:28:17.673147"} |
| {"step": 100, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.94, "mean_correctness_reward": 0.94, "mean_length_penalty": -0.099931640625, "mean_total_reward": 0.9100205078125, "std_total_reward": 0.23948497074610856, "total_tokens": 10233, "mean_completion_length": 204.66, "median_completion_length": 166, "min_completion_length": 96, "max_completion_length": 452, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.1835859569173654, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 4.402462959289551, "avg_step_time_sec": 22.66859637737274, "gpu_allocated_gb": 32.363235328, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2268.189811706543, "eta_sec": 2266.859637737274, "timestamp": "2026-02-15T13:28:24.914983"} |
| {"step": 101, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.76, "mean_correctness_reward": 0.76, "mean_length_penalty": -0.114326171875, "mean_total_reward": 0.7257021484375, "std_total_reward": 0.4413228388084308, "total_tokens": 11707, "mean_completion_length": 234.14, "median_completion_length": 185, "min_completion_length": 50, "max_completion_length": 796, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.1986049532821696, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 5.612526893615723, "avg_step_time_sec": 22.499724402286038, "gpu_allocated_gb": 32.36343552, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2275.938731908798, "eta_sec": 2227.4727158263177, "timestamp": "2026-02-15T13:28:32.748861"} |
| {"step": 102, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.085146484375, "mean_total_reward": 0.7744560546875, "std_total_reward": 0.4069406053954535, "total_tokens": 8719, "mean_completion_length": 174.38, "median_completion_length": 170, "min_completion_length": 71, "max_completion_length": 348, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.003919962378166574, "frac_prompts_all_correct": 0.8, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 9.033427476882935, "avg_step_time_sec": 22.367701883409538, "gpu_allocated_gb": 32.36390656, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2284.9794113636017, "eta_sec": 2192.034784574135, "timestamp": "2026-02-15T13:28:39.019266"} |
| {"step": 103, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.36, "mean_correctness_reward": 0.36, "mean_length_penalty": -0.1599609375, "mean_total_reward": 0.31201171875, "std_total_reward": 0.48422739676688326, "total_tokens": 16380, "mean_completion_length": 327.6, "median_completion_length": 282, "min_completion_length": 79, "max_completion_length": 930, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.08909764602663203, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 4.964251279830933, "avg_step_time_sec": 22.198736343568967, "gpu_allocated_gb": 32.363354112, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2289.9512276649475, "eta_sec": 2153.27742532619, "timestamp": "2026-02-15T13:28:47.838593"} |
| {"step": 104, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.48, "mean_correctness_reward": 0.48, "mean_length_penalty": -0.090732421875, "mean_total_reward": 0.4527802734375, "std_total_reward": 0.5027043409560775, "total_tokens": 9291, "mean_completion_length": 185.82, "median_completion_length": 147, "min_completion_length": 76, "max_completion_length": 525, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.2721080180934051, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 10.595959424972534, "avg_step_time_sec": 22.087171180890156, "gpu_allocated_gb": 32.364092416, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2300.5552010536194, "eta_sec": 2120.368433365455, "timestamp": "2026-02-15T13:28:55.474382"} |
| {"step": 105, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.8, "mean_correctness_reward": 0.8, "mean_length_penalty": -0.07740234375, "mean_total_reward": 0.776779296875, "std_total_reward": 0.4083644014261182, "total_tokens": 7926, "mean_completion_length": 158.52, "median_completion_length": 125, "min_completion_length": 64, "max_completion_length": 565, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.19849100336048192, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 6.6205596923828125, "avg_step_time_sec": 21.93987011909485, "gpu_allocated_gb": 32.3634944, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2307.1830689907074, "eta_sec": 2084.2876613140106, "timestamp": "2026-02-15T13:29:02.242867"} |
| {"step": 106, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.52, "mean_correctness_reward": 0.52, "mean_length_penalty": -0.184287109375, "mean_total_reward": 0.4647138671875, "std_total_reward": 0.5346341823309319, "total_tokens": 18871, "mean_completion_length": 377.42, "median_completion_length": 193, "min_completion_length": 77, "max_completion_length": 1896, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.11389706810637046, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 6.339509725570679, "avg_step_time_sec": 21.792696907835186, "gpu_allocated_gb": 32.363556352, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2313.5287976264954, "eta_sec": 2048.5135093365075, "timestamp": "2026-02-15T13:29:17.595306"} |
| {"step": 107, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.48, "mean_correctness_reward": 0.48, "mean_length_penalty": -0.0740625, "mean_total_reward": 0.45778125000000003, "std_total_reward": 0.5018533069449571, "total_tokens": 7584, "mean_completion_length": 151.68, "median_completion_length": 140, "min_completion_length": 60, "max_completion_length": 315, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.10064321558602075, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 20.813090085983276, "avg_step_time_sec": 21.783541703892645, "gpu_allocated_gb": 32.365500928, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2334.3489067554474, "eta_sec": 2025.869378462016, "timestamp": "2026-02-15T13:29:27.992370"} |
| {"step": 108, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.68, "mean_correctness_reward": 0.68, "mean_length_penalty": -0.080283203125, "mean_total_reward": 0.6559150390625, "std_total_reward": 0.4800800846774375, "total_tokens": 8221, "mean_completion_length": 164.42, "median_completion_length": 112, "min_completion_length": 67, "max_completion_length": 1066, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.23949293092718668, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 4.313328981399536, "avg_step_time_sec": 21.62178047498067, "gpu_allocated_gb": 32.363231232, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2338.6680331230164, "eta_sec": 1989.2038036982217, "timestamp": "2026-02-15T13:29:36.541203"} |
| {"step": 109, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.9, "mean_correctness_reward": 0.9, "mean_length_penalty": -0.059365234375, "mean_total_reward": 0.8821904296874999, "std_total_reward": 0.29994092849422144, "total_tokens": 6079, "mean_completion_length": 121.58, "median_completion_length": 126, "min_completion_length": 42, "max_completion_length": 224, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.22060113277527899, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 10.889412641525269, "avg_step_time_sec": 21.523318384765485, "gpu_allocated_gb": 32.364245504, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2349.5635669231415, "eta_sec": 1958.621973013659, "timestamp": "2026-02-15T13:29:42.577045"} |
| {"step": 110, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.72, "mean_correctness_reward": 0.72, "mean_length_penalty": -0.048251953125, "mean_total_reward": 0.7055244140625, "std_total_reward": 0.44953289924136064, "total_tokens": 4941, "mean_completion_length": 98.82, "median_completion_length": 89, "min_completion_length": 33, "max_completion_length": 285, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.33545812156834875, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 3.202845573425293, "avg_step_time_sec": 21.356768631935118, "gpu_allocated_gb": 32.36307456, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2352.772395849228, "eta_sec": 1922.1091768741605, "timestamp": "2026-02-15T13:29:46.050659"} |
| {"step": 111, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.38, "mean_correctness_reward": 0.38, "mean_length_penalty": -0.06654296875, "mean_total_reward": 0.360037109375, "std_total_reward": 0.4890769387119248, "total_tokens": 6814, "mean_completion_length": 136.28, "median_completion_length": 95, "min_completion_length": 50, "max_completion_length": 578, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.14201472367518528, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 3.70225191116333, "avg_step_time_sec": 21.197718931747985, "gpu_allocated_gb": 32.363179008, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2356.4804751873016, "eta_sec": 1886.5969849255707, "timestamp": "2026-02-15T13:29:51.531746"} |
| {"step": 112, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.68, "mean_correctness_reward": 0.68, "mean_length_penalty": -0.03732421875, "mean_total_reward": 0.668802734375, "std_total_reward": 0.4686430650771606, "total_tokens": 3822, "mean_completion_length": 76.44, "median_completion_length": 73, "min_completion_length": 30, "max_completion_length": 159, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.16074598469500964, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 6.4729063510894775, "avg_step_time_sec": 21.06624739084925, "gpu_allocated_gb": 32.363588608, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2362.958948135376, "eta_sec": 1853.829770394734, "timestamp": "2026-02-15T13:29:55.401715"} |
| {"step": 113, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.68, "mean_correctness_reward": 0.68, "mean_length_penalty": -0.04880859375, "mean_total_reward": 0.665357421875, "std_total_reward": 0.46928905601738335, "total_tokens": 4998, "mean_completion_length": 99.96, "median_completion_length": 78, "min_completion_length": 46, "max_completion_length": 359, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.3293127565423018, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 2.623382329940796, "avg_step_time_sec": 20.903036195619972, "gpu_allocated_gb": 32.3630208, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2365.588225364685, "eta_sec": 1818.5641490189375, "timestamp": "2026-02-15T13:29:59.750070"} |
| {"step": 114, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.62, "mean_correctness_reward": 0.62, "mean_length_penalty": -0.03923828125, "mean_total_reward": 0.608228515625, "std_total_reward": 0.48668534285666293, "total_tokens": 4018, "mean_completion_length": 80.36, "median_completion_length": 75, "min_completion_length": 37, "max_completion_length": 176, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.25904626401688596, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 4.798015356063843, "avg_step_time_sec": 20.761764082992286, "gpu_allocated_gb": 32.36327168, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2370.3921093940735, "eta_sec": 1785.5117111373365, "timestamp": "2026-02-15T13:30:02.946464"} |
| {"step": 115, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.84, "mean_correctness_reward": 0.84, "mean_length_penalty": -0.038291015625, "mean_total_reward": 0.8285126953125, "std_total_reward": 0.3670484698001407, "total_tokens": 3921, "mean_completion_length": 78.42, "median_completion_length": 70, "min_completion_length": 33, "max_completion_length": 180, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.2582171171220476, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 2.6727263927459717, "avg_step_time_sec": 20.6044681030771, "gpu_allocated_gb": 32.363019776, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2373.0732822418213, "eta_sec": 1751.3797887615535, "timestamp": "2026-02-15T13:30:05.635662"} |
| {"step": 116, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.78, "mean_correctness_reward": 0.78, "mean_length_penalty": -0.0212890625, "mean_total_reward": 0.77361328125, "std_total_reward": 0.414804144913832, "total_tokens": 2180, "mean_completion_length": 43.6, "median_completion_length": 45, "min_completion_length": 20, "max_completion_length": 67, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06067883409609843, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 2.669093132019043, "avg_step_time_sec": 20.44985280160246, "gpu_allocated_gb": 32.363017728, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2375.7516746520996, "eta_sec": 1717.7876353346066, "timestamp": "2026-02-15T13:30:07.574023"} |
| {"step": 117, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.7, "mean_correctness_reward": 0.7, "mean_length_penalty": -0.035390625, "mean_total_reward": 0.6893828125, "std_total_reward": 0.46092526137420153, "total_tokens": 3624, "mean_completion_length": 72.48, "median_completion_length": 67, "min_completion_length": 22, "max_completion_length": 142, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.10087680532347124, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 1.893538236618042, "avg_step_time_sec": 20.291251822414562, "gpu_allocated_gb": 32.362848768, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2377.6534790992737, "eta_sec": 1684.1739012604087, "timestamp": "2026-02-15T13:30:10.083550"} |
| {"step": 118, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.52, "mean_correctness_reward": 0.52, "mean_length_penalty": -0.068251953125, "mean_total_reward": 0.4995244140625, "std_total_reward": 0.5108516245133727, "total_tokens": 6989, "mean_completion_length": 139.78, "median_completion_length": 77, "min_completion_length": 15, "max_completion_length": 1335, "frac_has_think": 1.0, "frac_has_boxed": 0.9, "mean_reward_std_per_prompt": 0.28478961236071865, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 2.6394004821777344, "avg_step_time_sec": 20.14165986190408, "gpu_allocated_gb": 32.363058176, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2380.3020033836365, "eta_sec": 1651.6161086761347, "timestamp": "2026-02-15T13:30:19.612994"} |
| {"step": 119, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.72, "mean_correctness_reward": 0.72, "mean_length_penalty": -0.027958984375, "mean_total_reward": 0.7116123046875, "std_total_reward": 0.44962661232233686, "total_tokens": 2863, "mean_completion_length": 57.26, "median_completion_length": 52, "min_completion_length": 26, "max_completion_length": 113, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.2579236781031777, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 13.612282037734985, "avg_step_time_sec": 20.086791140692576, "gpu_allocated_gb": 32.36469248, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2393.922659635544, "eta_sec": 1627.0300823960986, "timestamp": "2026-02-15T13:30:26.031513"} |
| {"step": 120, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.54, "mean_correctness_reward": 0.54, "mean_length_penalty": -0.0378125, "mean_total_reward": 0.52865625, "std_total_reward": 0.5025833820454105, "total_tokens": 3872, "mean_completion_length": 77.44, "median_completion_length": 60, "min_completion_length": 12, "max_completion_length": 258, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.21301084349456284, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 2.201974868774414, "avg_step_time_sec": 19.937751005093258, "gpu_allocated_gb": 32.36293376, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2396.132455587387, "eta_sec": 1595.0200804074607, "timestamp": "2026-02-15T13:30:29.232916"} |
| {"step": 121, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.5, "mean_correctness_reward": 0.5, "mean_length_penalty": -0.0305859375, "mean_total_reward": 0.49082421875, "std_total_reward": 0.5023349196483436, "total_tokens": 3132, "mean_completion_length": 62.64, "median_completion_length": 55, "min_completion_length": 19, "max_completion_length": 342, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.16011252520124522, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 3.6200687885284424, "avg_step_time_sec": 19.802894127270406, "gpu_allocated_gb": 32.363176448, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2399.7609498500824, "eta_sec": 1564.428636054362, "timestamp": "2026-02-15T13:30:33.158658"} |
| {"step": 122, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.051474609375, "mean_total_reward": 0.3845576171875, "std_total_reward": 0.4961404987864006, "total_tokens": 5271, "mean_completion_length": 105.42, "median_completion_length": 65, "min_completion_length": 32, "max_completion_length": 1630, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.13393880294350108, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 4.072339773178101, "avg_step_time_sec": 19.673955157154897, "gpu_allocated_gb": 32.363281408, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2403.839103460312, "eta_sec": 1534.568502258082, "timestamp": "2026-02-15T13:30:44.529353"} |
| {"step": 123, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.56, "mean_correctness_reward": 0.56, "mean_length_penalty": -0.020849609375, "mean_total_reward": 0.5537451171875, "std_total_reward": 0.4978397538889936, "total_tokens": 2135, "mean_completion_length": 42.7, "median_completion_length": 38, "min_completion_length": 10, "max_completion_length": 103, "frac_has_think": 1.0, "frac_has_boxed": 0.96, "mean_reward_std_per_prompt": 0.23249587707597924, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 15.98195195198059, "avg_step_time_sec": 19.643938870933965, "gpu_allocated_gb": 32.36505856, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2419.8267500400543, "eta_sec": 1512.5832930619154, "timestamp": "2026-02-15T13:30:51.826222"} |
| {"step": 124, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.5, "mean_correctness_reward": 0.5, "mean_length_penalty": -0.021005859375, "mean_total_reward": 0.4936982421875, "std_total_reward": 0.5004545745518952, "total_tokens": 2151, "mean_completion_length": 43.02, "median_completion_length": 41, "min_completion_length": 16, "max_completion_length": 89, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.35790246173487733, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 2.5469143390655518, "avg_step_time_sec": 19.506059640838252, "gpu_allocated_gb": 32.36290304, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2422.3794510364532, "eta_sec": 1482.4605327037073, "timestamp": "2026-02-15T13:30:54.285515"} |
| {"step": 125, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.5, "mean_correctness_reward": 0.5, "mean_length_penalty": -0.0298046875, "mean_total_reward": 0.49105859375, "std_total_reward": 0.5016186167512603, "total_tokens": 3052, "mean_completion_length": 61.04, "median_completion_length": 49, "min_completion_length": 17, "max_completion_length": 247, "frac_has_think": 1.0, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.1605204980499416, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.9502170085906982, "avg_step_time_sec": 19.365612899780274, "gpu_allocated_gb": 32.362889728, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2424.3391618728638, "eta_sec": 1452.4209674835206, "timestamp": "2026-02-15T13:30:57.280778"} |
| {"step": 126, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.48, "mean_correctness_reward": 0.48, "mean_length_penalty": -0.025986328125, "mean_total_reward": 0.4722041015625, "std_total_reward": 0.4989020247178044, "total_tokens": 2661, "mean_completion_length": 53.22, "median_completion_length": 48, "min_completion_length": 16, "max_completion_length": 155, "frac_has_think": 1.0, "frac_has_boxed": 0.98, "mean_reward_std_per_prompt": 0.4582461319638849, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 3.49236798286438, "avg_step_time_sec": 19.239634765519035, "gpu_allocated_gb": 32.363175424, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2427.837236404419, "eta_sec": 1423.7329726484086, "timestamp": "2026-02-15T13:31:00.132952"} |
| {"step": 127, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.68, "mean_correctness_reward": 0.68, "mean_length_penalty": -0.01759765625, "mean_total_reward": 0.674720703125, "std_total_reward": 0.4664393041239236, "total_tokens": 1802, "mean_completion_length": 36.04, "median_completion_length": 35, "min_completion_length": 14, "max_completion_length": 211, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.3035501174950008, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 2.4224202632904053, "avg_step_time_sec": 19.107215753690465, "gpu_allocated_gb": 32.362989056, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2430.2668097019196, "eta_sec": 1394.826750019404, "timestamp": "2026-02-15T13:31:02.769464"} |
| {"step": 128, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.42, "mean_correctness_reward": 0.42, "mean_length_penalty": -0.022373046875, "mean_total_reward": 0.4132880859375, "std_total_reward": 0.4951768491784712, "total_tokens": 2291, "mean_completion_length": 45.82, "median_completion_length": 34, "min_completion_length": 12, "max_completion_length": 150, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.2392177646290842, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 2.6147446632385254, "avg_step_time_sec": 18.97836832329631, "gpu_allocated_gb": 32.363052544, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2432.8911135196686, "eta_sec": 1366.4425192773342, "timestamp": "2026-02-15T13:31:05.143908"} |
| {"step": 129, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.72, "mean_correctness_reward": 0.72, "mean_length_penalty": -0.02033203125, "mean_total_reward": 0.713900390625, "std_total_reward": 0.4496595399648591, "total_tokens": 2082, "mean_completion_length": 41.64, "median_completion_length": 36, "min_completion_length": 17, "max_completion_length": 188, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.2317503955570357, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 2.310882806777954, "avg_step_time_sec": 18.84916300921477, "gpu_allocated_gb": 32.36299008, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2435.208377599716, "eta_sec": 1338.2905736542486, "timestamp": "2026-02-15T13:31:07.612593"} |
| {"step": 130, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.46, "mean_correctness_reward": 0.46, "mean_length_penalty": -0.023681640625, "mean_total_reward": 0.4528955078125, "std_total_reward": 0.5002736198619762, "total_tokens": 2425, "mean_completion_length": 48.5, "median_completion_length": 34, "min_completion_length": 12, "max_completion_length": 184, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.25177232824411006, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 2.4960103034973145, "avg_step_time_sec": 18.7233695268631, "gpu_allocated_gb": 32.363037696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2437.709985256195, "eta_sec": 1310.6358668804169, "timestamp": "2026-02-15T13:31:10.159350"} |
| {"step": 131, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.46, "mean_correctness_reward": 0.46, "mean_length_penalty": -0.01814453125, "mean_total_reward": 0.454556640625, "std_total_reward": 0.49769423047355227, "total_tokens": 1858, "mean_completion_length": 37.16, "median_completion_length": 31, "min_completion_length": 14, "max_completion_length": 96, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.2517812405761982, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 2.8188586235046387, "avg_step_time_sec": 18.60196104668479, "gpu_allocated_gb": 32.3630336, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2440.53613448143, "eta_sec": 1283.5353122212505, "timestamp": "2026-02-15T13:31:12.442439"} |
| {"step": 132, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.44, "mean_correctness_reward": 0.44, "mean_length_penalty": -0.01619140625, "mean_total_reward": 0.435142578125, "std_total_reward": 0.4962773150788922, "total_tokens": 1658, "mean_completion_length": 33.16, "median_completion_length": 27, "min_completion_length": 18, "max_completion_length": 88, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.33745918879590525, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 1.940507411956787, "avg_step_time_sec": 18.475737913088366, "gpu_allocated_gb": 32.362873856, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2442.4848766326904, "eta_sec": 1256.350178090009, "timestamp": "2026-02-15T13:31:14.353713"} |
| {"step": 133, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.01224609375, "mean_total_reward": 0.596326171875, "std_total_reward": 0.48948977926336307, "total_tokens": 1254, "mean_completion_length": 25.08, "median_completion_length": 19, "min_completion_length": 15, "max_completion_length": 58, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.19998127747173525, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 1.928213357925415, "avg_step_time_sec": 18.351320435229997, "gpu_allocated_gb": 32.362906112, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2444.421983718872, "eta_sec": 1229.5384691604097, "timestamp": "2026-02-15T13:31:16.152716"} |
| {"step": 134, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.14, "mean_correctness_reward": 0.14, "mean_length_penalty": -0.0164453125, "mean_total_reward": 0.13506640625, "std_total_reward": 0.34727836599038936, "total_tokens": 1684, "mean_completion_length": 33.68, "median_completion_length": 22, "min_completion_length": 18, "max_completion_length": 250, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.1598690166327582, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.8377666473388672, "avg_step_time_sec": 18.22808495920096, "gpu_allocated_gb": 32.36285696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2446.269766807556, "eta_sec": 1203.0536073072633, "timestamp": "2026-02-15T13:31:19.095415"} |
| {"step": 135, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.66, "mean_correctness_reward": 0.66, "mean_length_penalty": -0.0162890625, "mean_total_reward": 0.65511328125, "std_total_reward": 0.4735447929891132, "total_tokens": 1668, "mean_completion_length": 33.36, "median_completion_length": 32, "min_completion_length": 13, "max_completion_length": 65, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.29186564411358057, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 3.641237735748291, "avg_step_time_sec": 18.120034239027234, "gpu_allocated_gb": 32.36314112, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2449.921547651291, "eta_sec": 1177.8022255367703, "timestamp": "2026-02-15T13:31:21.646441"} |
| {"step": 136, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.56, "mean_correctness_reward": 0.56, "mean_length_penalty": -0.020244140625, "mean_total_reward": 0.5539267578125, "std_total_reward": 0.4979502584260036, "total_tokens": 2073, "mean_completion_length": 41.46, "median_completion_length": 41, "min_completion_length": 16, "max_completion_length": 70, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.08079400640232642, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.817880630493164, "avg_step_time_sec": 18.000165462493896, "gpu_allocated_gb": 32.362840064, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2451.7473237514496, "eta_sec": 1152.0105895996094, "timestamp": "2026-02-15T13:31:23.535247"} |
| {"step": 137, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.54, "mean_correctness_reward": 0.54, "mean_length_penalty": -0.0188671875, "mean_total_reward": 0.53433984375, "std_total_reward": 0.4990439082992432, "total_tokens": 1932, "mean_completion_length": 38.64, "median_completion_length": 34, "min_completion_length": 18, "max_completion_length": 188, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.14224657805792593, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.8574578762054443, "avg_step_time_sec": 17.88233548011223, "gpu_allocated_gb": 32.362866688, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2453.6109499931335, "eta_sec": 1126.5871352470704, "timestamp": "2026-02-15T13:31:26.062195"} |
| {"step": 138, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.2, "mean_correctness_reward": 0.2, "mean_length_penalty": -0.018408203125, "mean_total_reward": 0.19447753906249998, "std_total_reward": 0.40003937959542046, "total_tokens": 1885, "mean_completion_length": 37.7, "median_completion_length": 25, "min_completion_length": 16, "max_completion_length": 445, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.11986806770446505, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 2.5660362243652344, "avg_step_time_sec": 17.771347804345947, "gpu_allocated_gb": 32.363042816, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2456.185232400894, "eta_sec": 1101.8235638694487, "timestamp": "2026-02-15T13:31:30.019072"} |
| {"step": 139, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.5, "mean_correctness_reward": 0.5, "mean_length_penalty": -0.016826171875, "mean_total_reward": 0.4949521484375, "std_total_reward": 0.5008120635114182, "total_tokens": 1723, "mean_completion_length": 34.46, "median_completion_length": 32, "min_completion_length": 15, "max_completion_length": 104, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.2924280394720516, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 4.830698251724243, "avg_step_time_sec": 17.678249606125647, "gpu_allocated_gb": 32.36340736, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2461.0214791297913, "eta_sec": 1078.3732259736644, "timestamp": "2026-02-15T13:31:33.007902"} |
| {"step": 140, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.5, "mean_correctness_reward": 0.5, "mean_length_penalty": -0.021689453125, "mean_total_reward": 0.4934931640625, "std_total_reward": 0.49990904675165365, "total_tokens": 2221, "mean_completion_length": 44.42, "median_completion_length": 35, "min_completion_length": 10, "max_completion_length": 100, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.15879128108179688, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 2.201413631439209, "avg_step_time_sec": 17.56770077773503, "gpu_allocated_gb": 32.362958336, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2463.2301194667816, "eta_sec": 1054.062046664102, "timestamp": "2026-02-15T13:31:35.225988"} |
| {"step": 141, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.44, "mean_correctness_reward": 0.44, "mean_length_penalty": -0.01458984375, "mean_total_reward": 0.43562304687500003, "std_total_reward": 0.49673182920292114, "total_tokens": 1494, "mean_completion_length": 29.88, "median_completion_length": 22, "min_completion_length": 10, "max_completion_length": 76, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.26409596851167433, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 2.0502331256866455, "avg_step_time_sec": 17.457647815663766, "gpu_allocated_gb": 32.362918912, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2465.2890207767487, "eta_sec": 1030.0012211241622, "timestamp": "2026-02-15T13:31:37.071854"} |
| {"step": 142, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.46, "mean_correctness_reward": 0.46, "mean_length_penalty": -0.01720703125, "mean_total_reward": 0.454837890625, "std_total_reward": 0.49933807559662374, "total_tokens": 1762, "mean_completion_length": 35.24, "median_completion_length": 28, "min_completion_length": 10, "max_completion_length": 92, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0923040040140957, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.783341646194458, "avg_step_time_sec": 17.3472653778506, "gpu_allocated_gb": 32.362877952, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2467.0801923274994, "eta_sec": 1006.1413919153348, "timestamp": "2026-02-15T13:31:38.960243"} |
| {"step": 143, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.26, "mean_correctness_reward": 0.26, "mean_length_penalty": -0.013701171875, "mean_total_reward": 0.2558896484375, "std_total_reward": 0.4389274124242168, "total_tokens": 1403, "mean_completion_length": 28.06, "median_completion_length": 22, "min_completion_length": 10, "max_completion_length": 92, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.1808359099055104, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 2.030667304992676, "avg_step_time_sec": 17.240156300418025, "gpu_allocated_gb": 32.36288512, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2469.1183738708496, "eta_sec": 982.6889091238274, "timestamp": "2026-02-15T13:31:40.998572"} |
| {"step": 144, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.48, "mean_correctness_reward": 0.48, "mean_length_penalty": -0.01013671875, "mean_total_reward": 0.476958984375, "std_total_reward": 0.5002473333536475, "total_tokens": 1038, "mean_completion_length": 20.76, "median_completion_length": 21, "min_completion_length": 10, "max_completion_length": 35, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.09863502978799003, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.8986921310424805, "avg_step_time_sec": 17.133618354797363, "gpu_allocated_gb": 32.362909696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2471.0259487628937, "eta_sec": 959.4826278686523, "timestamp": "2026-02-15T13:31:42.564488"} |
| {"step": 145, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.56, "mean_correctness_reward": 0.56, "mean_length_penalty": -0.007001953125, "mean_total_reward": 0.5578994140625, "std_total_reward": 0.496290519779112, "total_tokens": 717, "mean_completion_length": 14.34, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 37, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.12047080538180008, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.5438945293426514, "avg_step_time_sec": 17.02610301807009, "gpu_allocated_gb": 32.362812416, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2472.577520608902, "eta_sec": 936.4356659938549, "timestamp": "2026-02-15T13:31:44.104185"} |
| {"step": 146, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.34, "mean_correctness_reward": 0.34, "mean_length_penalty": -0.0069140625, "mean_total_reward": 0.33792578125, "std_total_reward": 0.47351555682913327, "total_tokens": 708, "mean_completion_length": 14.16, "median_completion_length": 16, "min_completion_length": 10, "max_completion_length": 29, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.09183289928527338, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.50295090675354, "avg_step_time_sec": 16.91978005840354, "gpu_allocated_gb": 32.362842624, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2474.086715221405, "eta_sec": 913.6681231537912, "timestamp": "2026-02-15T13:31:45.547462"} |
| {"step": 147, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.2, "mean_correctness_reward": 0.2, "mean_length_penalty": -0.005478515625, "mean_total_reward": 0.19835644531250002, "std_total_reward": 0.4000161315061257, "total_tokens": 561, "mean_completion_length": 11.22, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 14, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 3.808593750000001e-05, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.9482421875, "avg_step_time_sec": 16.817932862002834, "gpu_allocated_gb": 32.36281344, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2476.041632413864, "eta_sec": 891.3504416861502, "timestamp": "2026-02-15T13:31:47.442637"} |
| {"step": 148, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.18, "mean_correctness_reward": 0.18, "mean_length_penalty": -0.006845703125, "mean_total_reward": 0.1779462890625, "std_total_reward": 0.38439519610310885, "total_tokens": 701, "mean_completion_length": 14.02, "median_completion_length": 14, "min_completion_length": 11, "max_completion_length": 36, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06038750550117614, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.3671433925628662, "avg_step_time_sec": 16.71353563585797, "gpu_allocated_gb": 32.36279808, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2477.4169659614563, "eta_sec": 869.1038530646144, "timestamp": "2026-02-15T13:31:48.918184"} |
| {"step": 149, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.56, "mean_correctness_reward": 0.56, "mean_length_penalty": -0.005615234375, "mean_total_reward": 0.5583154296875, "std_total_reward": 0.49645794163169676, "total_tokens": 575, "mean_completion_length": 11.5, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 30, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.15151137274107945, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 1.5334436893463135, "avg_step_time_sec": 16.611655824136413, "gpu_allocated_gb": 32.362842624, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2478.9558522701263, "eta_sec": 847.1944470309571, "timestamp": "2026-02-15T13:31:50.400826"} |
| {"step": 150, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.005576171875, "mean_total_reward": 0.3983271484375, "std_total_reward": 0.4899482603209047, "total_tokens": 571, "mean_completion_length": 11.42, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 20, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.00010595580440494337, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.4308490753173828, "avg_step_time_sec": 16.510450445810953, "gpu_allocated_gb": 32.362779648, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2480.3925235271454, "eta_sec": 825.5225222905476, "timestamp": "2026-02-15T13:31:52.493643"} |
| {"step": 151, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.74, "mean_correctness_reward": 0.74, "mean_length_penalty": -0.00560546875, "mean_total_reward": 0.738318359375, "std_total_reward": 0.4386527887403335, "total_tokens": 574, "mean_completion_length": 11.48, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 24, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.23810263658632713, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.0, "step_time_sec": 1.4877233505249023, "avg_step_time_sec": 16.410962186901774, "gpu_allocated_gb": 32.362793984, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2482.4967045783997, "eta_sec": 804.137147158187, "timestamp": "2026-02-15T13:31:53.954082"} |
| {"step": 152, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.12, "mean_correctness_reward": 0.12, "mean_length_penalty": -0.005615234375, "mean_total_reward": 0.1183154296875, "std_total_reward": 0.32504275689938356, "total_tokens": 575, "mean_completion_length": 11.5, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 15, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.2316769054851842, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.4389605522155762, "avg_step_time_sec": 16.31246217614726, "gpu_allocated_gb": 32.362863616, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2483.941396713257, "eta_sec": 782.9981844550684, "timestamp": "2026-02-15T13:31:55.337266"} |
| {"step": 153, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.34, "mean_correctness_reward": 0.34, "mean_length_penalty": -0.00654296875, "mean_total_reward": 0.338037109375, "std_total_reward": 0.47400497268531017, "total_tokens": 670, "mean_completion_length": 13.4, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 25, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.09169516325307842, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.3667194843292236, "avg_step_time_sec": 16.21477758339028, "gpu_allocated_gb": 32.362800128, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2485.316578388214, "eta_sec": 762.0945464193431, "timestamp": "2026-02-15T13:31:56.803458"} |
| {"step": 154, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.2, "mean_correctness_reward": 0.2, "mean_length_penalty": -0.005693359375, "mean_total_reward": 0.19829199218750002, "std_total_reward": 0.4000483536688665, "total_tokens": 583, "mean_completion_length": 11.66, "median_completion_length": 12, "min_completion_length": 10, "max_completion_length": 13, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 5.566406249999998e-05, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.4659154415130615, "avg_step_time_sec": 16.11900575130017, "gpu_allocated_gb": 32.362857472, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2486.790782928467, "eta_sec": 741.4742645598077, "timestamp": "2026-02-15T13:31:58.180915"} |
| {"step": 155, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.36, "mean_correctness_reward": 0.36, "mean_length_penalty": -0.00501953125, "mean_total_reward": 0.358494140625, "std_total_reward": 0.48003076523892374, "total_tokens": 514, "mean_completion_length": 10.28, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 11, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.12001435247896161, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.3855226039886475, "avg_step_time_sec": 16.02395102131751, "gpu_allocated_gb": 32.362800128, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2488.1847007274628, "eta_sec": 721.077795959288, "timestamp": "2026-02-15T13:31:59.557096"} |
| {"step": 156, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.22, "mean_correctness_reward": 0.22, "mean_length_penalty": -0.005537109375, "mean_total_reward": 0.2183388671875, "std_total_reward": 0.4143435204851467, "total_tokens": 567, "mean_completion_length": 11.34, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 14, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0600263671875, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.3580565452575684, "avg_step_time_sec": 15.929938877240206, "gpu_allocated_gb": 32.362818048, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2489.5485095977783, "eta_sec": 700.9173105985691, "timestamp": "2026-02-15T13:32:00.960744"} |
| {"step": 157, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.28, "mean_correctness_reward": 0.28, "mean_length_penalty": -0.005048828125, "mean_total_reward": 0.2784853515625, "std_total_reward": 0.44902995033561904, "total_tokens": 517, "mean_completion_length": 10.34, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 12, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.18332693663033964, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.4248690605163574, "avg_step_time_sec": 15.83754989751585, "gpu_allocated_gb": 32.362802176, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2490.980992794037, "eta_sec": 681.0146455931815, "timestamp": "2026-02-15T13:32:02.377776"} |
| {"step": 158, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.12, "mean_correctness_reward": 0.12, "mean_length_penalty": -0.006396484375, "mean_total_reward": 0.1180810546875, "std_total_reward": 0.3251295747314113, "total_tokens": 655, "mean_completion_length": 13.1, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 21, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.09812287750178122, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.3767180442810059, "avg_step_time_sec": 15.746025645280186, "gpu_allocated_gb": 32.36283136, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2492.36620926857, "eta_sec": 661.3330771017678, "timestamp": "2026-02-15T13:32:03.797480"} |
| {"step": 159, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.12, "mean_correctness_reward": 0.12, "mean_length_penalty": -0.00560546875, "mean_total_reward": 0.118318359375, "std_total_reward": 0.3250147782941864, "total_tokens": 574, "mean_completion_length": 11.48, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 18, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.2317569660046236, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.4142403602600098, "avg_step_time_sec": 15.65588863090899, "gpu_allocated_gb": 32.362838528, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2493.786575317383, "eta_sec": 641.8914338672686, "timestamp": "2026-02-15T13:32:05.192073"} |
| {"step": 160, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.46, "mean_correctness_reward": 0.46, "mean_length_penalty": -0.005078125, "mean_total_reward": 0.4584765625, "std_total_reward": 0.49844563597732633, "total_tokens": 520, "mean_completion_length": 10.4, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 11, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.14, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 1.87916898727417, "avg_step_time_sec": 15.569784133136272, "gpu_allocated_gb": 32.362806272, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2495.671452522278, "eta_sec": 622.7913653254509, "timestamp": "2026-02-15T13:32:07.060479"} |
| {"step": 161, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.0, "mean_correctness_reward": 0.0, "mean_length_penalty": -0.0065625, "mean_total_reward": -0.00196875, "std_total_reward": 0.0007538809500288531, "total_tokens": 672, "mean_completion_length": 13.44, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 24, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 5.510815334738274e-05, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 1.0, "step_time_sec": 1.377328634262085, "avg_step_time_sec": 15.481632235627737, "gpu_allocated_gb": 32.362833408, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2497.0570425987244, "eta_sec": 603.7836571894818, "timestamp": "2026-02-15T13:32:08.521066"} |
| {"step": 162, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.04, "mean_correctness_reward": 0.04, "mean_length_penalty": -0.00576171875, "mean_total_reward": 0.038271484375000005, "std_total_reward": 0.19596822397377103, "total_tokens": 590, "mean_completion_length": 11.8, "median_completion_length": 11, "min_completion_length": 11, "max_completion_length": 14, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.12000000000000001, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.4555766582489014, "avg_step_time_sec": 15.395051645643917, "gpu_allocated_gb": 32.362785792, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2498.5194039344788, "eta_sec": 585.0119625344689, "timestamp": "2026-02-15T13:32:09.901182"} |
| {"step": 163, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.34, "mean_correctness_reward": 0.34, "mean_length_penalty": -0.006240234375, "mean_total_reward": 0.3381279296875, "std_total_reward": 0.47387763193959787, "total_tokens": 639, "mean_completion_length": 12.78, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 21, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.09173158314554206, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.3625860214233398, "avg_step_time_sec": 15.308962899483054, "gpu_allocated_gb": 32.36281344, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2499.887759923935, "eta_sec": 566.431627280873, "timestamp": "2026-02-15T13:32:11.303249"} |
| {"step": 164, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.1, "mean_correctness_reward": 0.1, "mean_length_penalty": -0.006279296875, "mean_total_reward": 0.0981162109375, "std_total_reward": 0.30004251561739825, "total_tokens": 643, "mean_completion_length": 12.86, "median_completion_length": 12, "min_completion_length": 10, "max_completion_length": 24, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.1002069623651014, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.4469695091247559, "avg_step_time_sec": 15.224438549541846, "gpu_allocated_gb": 32.362789376, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2501.34064078331, "eta_sec": 548.0797877835065, "timestamp": "2026-02-15T13:32:12.810104"} |
| {"step": 165, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.38, "mean_correctness_reward": 0.38, "mean_length_penalty": -0.00623046875, "mean_total_reward": 0.378130859375, "std_total_reward": 0.4853530231798631, "total_tokens": 638, "mean_completion_length": 12.76, "median_completion_length": 12, "min_completion_length": 10, "max_completion_length": 17, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.059982421875000004, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.4798545837402344, "avg_step_time_sec": 15.1411380406582, "gpu_allocated_gb": 32.36284928, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2502.8265664577484, "eta_sec": 529.939831423037, "timestamp": "2026-02-15T13:32:14.220113"} |
| {"step": 166, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.32, "mean_correctness_reward": 0.32, "mean_length_penalty": -0.0091796875, "mean_total_reward": 0.31724609375, "std_total_reward": 0.4672015878856143, "total_tokens": 940, "mean_completion_length": 18.8, "median_completion_length": 12, "min_completion_length": 11, "max_completion_length": 100, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0991032472170426, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.3548212051391602, "avg_step_time_sec": 15.05808793923941, "gpu_allocated_gb": 32.362781184, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2504.188958644867, "eta_sec": 511.97498993413996, "timestamp": "2026-02-15T13:32:16.074308"} |
| {"step": 167, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.64, "mean_correctness_reward": 0.64, "mean_length_penalty": -0.00830078125, "mean_total_reward": 0.6375097656250001, "std_total_reward": 0.4812406388400307, "total_tokens": 850, "mean_completion_length": 17, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 45, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.08012010487552153, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.2, "step_time_sec": 1.974463701248169, "avg_step_time_sec": 14.979742883922098, "gpu_allocated_gb": 32.3629568, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2506.169613838196, "eta_sec": 494.3315151694292, "timestamp": "2026-02-15T13:32:17.752712"} |
| {"step": 168, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.04, "mean_correctness_reward": 0.04, "mean_length_penalty": -0.0073046875, "mean_total_reward": 0.03780859375, "std_total_reward": 0.19611143467161715, "total_tokens": 748, "mean_completion_length": 14.96, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 32, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.08001904397452313, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.5934178829193115, "avg_step_time_sec": 14.900062377963748, "gpu_allocated_gb": 32.36287488, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2507.76856136322, "eta_sec": 476.8019960948399, "timestamp": "2026-02-15T13:32:19.244851"} |
| {"step": 169, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.2, "mean_correctness_reward": 0.2, "mean_length_penalty": -0.00662109375, "mean_total_reward": 0.198013671875, "std_total_reward": 0.4001879924370189, "total_tokens": 678, "mean_completion_length": 13.56, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 23, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 3.841209424784827e-05, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.4560067653656006, "avg_step_time_sec": 14.820511753037131, "gpu_allocated_gb": 32.362809856, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2509.2296895980835, "eta_sec": 459.43586434415107, "timestamp": "2026-02-15T13:32:20.660559"} |
| {"step": 170, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.38, "mean_correctness_reward": 0.38, "mean_length_penalty": -0.00517578125, "mean_total_reward": 0.378447265625, "std_total_reward": 0.4854552788610377, "total_tokens": 530, "mean_completion_length": 10.6, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 13, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.19797958971132712, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.479684591293335, "avg_step_time_sec": 14.742036299144521, "gpu_allocated_gb": 32.362783232, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2510.7154829502106, "eta_sec": 442.2610889743356, "timestamp": "2026-02-15T13:32:22.078453"} |
| {"step": 171, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.006669921875, "mean_total_reward": 0.5979990234375, "std_total_reward": 0.4899566014679868, "total_tokens": 683, "mean_completion_length": 13.66, "median_completion_length": 14, "min_completion_length": 11, "max_completion_length": 16, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 2.0507812500000005e-05, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.343552589416504, "avg_step_time_sec": 14.663682593239677, "gpu_allocated_gb": 32.362785792, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2512.063782930374, "eta_sec": 425.2467952039506, "timestamp": "2026-02-15T13:32:23.501486"} |
| {"step": 172, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.005283203125, "mean_total_reward": 0.39841503906250003, "std_total_reward": 0.4899362323305159, "total_tokens": 541, "mean_completion_length": 10.82, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 12, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 8.789062499999994e-06, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.4565777778625488, "avg_step_time_sec": 14.58689710012702, "gpu_allocated_gb": 32.36285696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2513.527129650116, "eta_sec": 408.43311880355657, "timestamp": "2026-02-15T13:32:24.895199"} |
| {"step": 173, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.00619140625, "mean_total_reward": 0.398142578125, "std_total_reward": 0.4901588588189161, "total_tokens": 634, "mean_completion_length": 12.68, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 19, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 1.4352478961620176e-05, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.835141658782959, "avg_step_time_sec": 14.513187531101911, "gpu_allocated_gb": 32.3627904, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2515.3682193756104, "eta_sec": 391.8560633397516, "timestamp": "2026-02-15T13:32:26.794700"} |
| {"step": 174, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.2, "mean_correctness_reward": 0.2, "mean_length_penalty": -0.00591796875, "mean_total_reward": 0.198224609375, "std_total_reward": 0.40015543060755077, "total_tokens": 606, "mean_completion_length": 12.12, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 18, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 7.557995792324033e-05, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.4058289527893066, "avg_step_time_sec": 14.437857884100113, "gpu_allocated_gb": 32.362780672, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2516.7813107967377, "eta_sec": 375.38430498660296, "timestamp": "2026-02-15T13:32:28.183266"} |
| {"step": 175, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.2, "mean_correctness_reward": 0.2, "mean_length_penalty": -0.0061328125, "mean_total_reward": 0.19816015625, "std_total_reward": 0.40004117461001665, "total_tokens": 628, "mean_completion_length": 12.56, "median_completion_length": 12, "min_completion_length": 10, "max_completion_length": 16, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 1.1718749999999993e-05, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.3869900703430176, "avg_step_time_sec": 14.363281496592931, "gpu_allocated_gb": 32.362796032, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2518.1742975711823, "eta_sec": 359.0820374148233, "timestamp": "2026-02-15T13:32:29.576467"} |
| {"step": 176, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.3, "mean_correctness_reward": 0.3, "mean_length_penalty": -0.00529296875, "mean_total_reward": 0.298412109375, "std_total_reward": 0.45824225703564014, "total_tokens": 542, "mean_completion_length": 10.84, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 13, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.10001171875, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.3760571479797363, "avg_step_time_sec": 14.289490449157627, "gpu_allocated_gb": 32.36279552, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2519.5585079193115, "eta_sec": 342.94777077978307, "timestamp": "2026-02-15T13:32:30.936987"} |
| {"step": 177, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.6, "mean_correctness_reward": 0.6, "mean_length_penalty": -0.0057421875, "mean_total_reward": 0.59827734375, "std_total_reward": 0.4900942074501682, "total_tokens": 588, "mean_completion_length": 11.76, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 18, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 2.8704957923240352e-05, "frac_prompts_all_correct": 0.6, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.3544948101043701, "avg_step_time_sec": 14.216411377750548, "gpu_allocated_gb": 32.362768384, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2520.920572042465, "eta_sec": 326.9774616882626, "timestamp": "2026-02-15T13:32:32.315070"} |
| {"step": 178, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.0, "mean_correctness_reward": 0.0, "mean_length_penalty": -0.006083984375, "mean_total_reward": -0.0018251953125, "std_total_reward": 0.00040192280201767393, "total_tokens": 623, "mean_completion_length": 12.46, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 19, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 8.197263279373124e-05, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 1.0, "step_time_sec": 1.3660690784454346, "avg_step_time_sec": 14.144218443484789, "gpu_allocated_gb": 32.3628032, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2522.2924766540527, "eta_sec": 311.17280575666535, "timestamp": "2026-02-15T13:32:33.696679"} |
| {"step": 179, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.12, "mean_correctness_reward": 0.12, "mean_length_penalty": -0.00650390625, "mean_total_reward": 0.118048828125, "std_total_reward": 0.32512373267326056, "total_tokens": 666, "mean_completion_length": 13.32, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 24, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.17804891849663967, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.3751230239868164, "avg_step_time_sec": 14.07288271488424, "gpu_allocated_gb": 32.362806272, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2523.6737444400787, "eta_sec": 295.530537012569, "timestamp": "2026-02-15T13:32:35.182175"} |
| {"step": 180, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.24, "mean_correctness_reward": 0.24, "mean_length_penalty": -0.005517578125, "mean_total_reward": 0.2383447265625, "std_total_reward": 0.42710797245625, "total_tokens": 565, "mean_completion_length": 11.3, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 20, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.08014466574277589, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.6103532314300537, "avg_step_time_sec": 14.003646439976162, "gpu_allocated_gb": 32.362865664, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2525.290164709091, "eta_sec": 280.07292879952325, "timestamp": "2026-02-15T13:32:36.730815"} |
| {"step": 181, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.18, "mean_correctness_reward": 0.18, "mean_length_penalty": -0.005712890625, "mean_total_reward": 0.1782861328125, "std_total_reward": 0.3842355535679234, "total_tokens": 585, "mean_completion_length": 11.7, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 14, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06009301091919395, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.502859354019165, "avg_step_time_sec": 13.9345813179543, "gpu_allocated_gb": 32.362815488, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2526.8024973869324, "eta_sec": 264.75704504113173, "timestamp": "2026-02-15T13:32:38.209744"} |
| {"step": 182, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.54, "mean_correctness_reward": 0.54, "mean_length_penalty": -0.00580078125, "mean_total_reward": 0.538259765625, "std_total_reward": 0.49834914073347586, "total_tokens": 594, "mean_completion_length": 11.88, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 17, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0916808107741168, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.5245606899261475, "avg_step_time_sec": 13.866394391426674, "gpu_allocated_gb": 32.362783744, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2528.3343517780304, "eta_sec": 249.59509904568012, "timestamp": "2026-02-15T13:32:39.763333"} |
| {"step": 183, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.0, "mean_correctness_reward": 0.0, "mean_length_penalty": -0.005361328125, "mean_total_reward": -0.0016083984375, "std_total_reward": 0.00016177075420638893, "total_tokens": 549, "mean_completion_length": 10.98, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 13, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 2.7777993692936104e-05, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 1.0, "step_time_sec": 1.4786334037780762, "avg_step_time_sec": 13.798701708434058, "gpu_allocated_gb": 32.362777088, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2529.820374727249, "eta_sec": 234.577929043379, "timestamp": "2026-02-15T13:32:41.232486"} |
| {"step": 184, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.0, "mean_correctness_reward": 0.0, "mean_length_penalty": -0.00841796875, "mean_total_reward": -0.002525390625, "std_total_reward": 0.0014865270173266907, "total_tokens": 862, "mean_completion_length": 17.24, "median_completion_length": 12, "min_completion_length": 11, "max_completion_length": 42, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.00020336446711421237, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 1.0, "step_time_sec": 1.51005220413208, "avg_step_time_sec": 13.731915569823721, "gpu_allocated_gb": 32.362788864, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2531.3382012844086, "eta_sec": 219.71064911717954, "timestamp": "2026-02-15T13:32:42.920398"} |
| {"step": 185, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.18, "mean_correctness_reward": 0.18, "mean_length_penalty": -0.00583984375, "mean_total_reward": 0.178248046875, "std_total_reward": 0.38425344939411654, "total_tokens": 598, "mean_completion_length": 11.96, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 16, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.06001171875, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 2.0739057064056396, "avg_step_time_sec": 13.66889930029173, "gpu_allocated_gb": 32.36285696, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2533.4215993881226, "eta_sec": 205.03348950437595, "timestamp": "2026-02-15T13:32:44.874355"} |
| {"step": 186, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.2, "mean_correctness_reward": 0.2, "mean_length_penalty": -0.005498046875, "mean_total_reward": 0.1983505859375, "std_total_reward": 0.39987257886429683, "total_tokens": 563, "mean_completion_length": 11.26, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 13, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.18328960228350227, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.5133767127990723, "avg_step_time_sec": 13.603547028315965, "gpu_allocated_gb": 32.36285184, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2534.9423689842224, "eta_sec": 190.4496583964235, "timestamp": "2026-02-15T13:32:46.331772"} |
| {"step": 187, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.005703125, "mean_total_reward": 0.3982890625, "std_total_reward": 0.4897102275664506, "total_tokens": 584, "mean_completion_length": 11.68, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 16, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.199947265625, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.4, "step_time_sec": 1.3625478744506836, "avg_step_time_sec": 13.538087139792621, "gpu_allocated_gb": 32.362806272, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2536.3119657039642, "eta_sec": 175.99513281730407, "timestamp": "2026-02-15T13:32:47.764022"} |
| {"step": 188, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.26, "mean_correctness_reward": 0.26, "mean_length_penalty": -0.005908203125, "mean_total_reward": 0.2582275390625, "std_total_reward": 0.43859628807776746, "total_tokens": 605, "mean_completion_length": 12.1, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 18, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.091693059554361, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.4245860576629639, "avg_step_time_sec": 13.473653623398315, "gpu_allocated_gb": 32.362863104, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2537.7444887161255, "eta_sec": 161.68384348077979, "timestamp": "2026-02-15T13:32:49.165284"} |
| {"step": 189, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.2, "mean_correctness_reward": 0.2, "mean_length_penalty": -0.00525390625, "mean_total_reward": 0.198423828125, "std_total_reward": 0.3999091849994799, "total_tokens": 538, "mean_completion_length": 10.76, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 12, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 1.1718749999999993e-05, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.40775728225708, "avg_step_time_sec": 13.409812902016615, "gpu_allocated_gb": 32.362817536, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2539.1591577529907, "eta_sec": 147.50794192218277, "timestamp": "2026-02-15T13:32:50.508151"} |
| {"step": 190, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.2, "mean_correctness_reward": 0.2, "mean_length_penalty": -0.005478515625, "mean_total_reward": 0.1983564453125, "std_total_reward": 0.4000894466300707, "total_tokens": 561, "mean_completion_length": 11.22, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 15, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 8.789062499999994e-06, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.3115978240966797, "avg_step_time_sec": 13.346138085817035, "gpu_allocated_gb": 32.362768896, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2540.4767887592316, "eta_sec": 133.46138085817034, "timestamp": "2026-02-15T13:32:51.877675"} |
| {"step": 191, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.0, "mean_correctness_reward": 0.0, "mean_length_penalty": -0.00529296875, "mean_total_reward": -0.001587890625, "std_total_reward": 0.00015324254098385176, "total_tokens": 542, "mean_completion_length": 10.84, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 13, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 4.409143248375261e-05, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 1.0, "step_time_sec": 1.380345106124878, "avg_step_time_sec": 13.283489955033307, "gpu_allocated_gb": 32.36279296, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2541.8653225898743, "eta_sec": 119.55140959529976, "timestamp": "2026-02-15T13:32:53.354305"} |
| {"step": 192, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.2, "mean_correctness_reward": 0.2, "mean_length_penalty": -0.00537109375, "mean_total_reward": 0.198388671875, "std_total_reward": 0.40007327839069107, "total_tokens": 550, "mean_completion_length": 11, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 13, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.4711189270019531, "avg_step_time_sec": 13.22196718926231, "gpu_allocated_gb": 32.362904576, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2543.3443717956543, "eta_sec": 105.77573751409848, "timestamp": "2026-02-15T13:32:54.734162"} |
| {"step": 193, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.34, "mean_correctness_reward": 0.34, "mean_length_penalty": -0.005224609375, "mean_total_reward": 0.3384326171875, "std_total_reward": 0.47361538881953574, "total_tokens": 535, "mean_completion_length": 10.7, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 12, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.09164980713438549, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.3948643207550049, "avg_step_time_sec": 13.16068686351875, "gpu_allocated_gb": 32.362776576, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2544.7476613521576, "eta_sec": 92.12480804463125, "timestamp": "2026-02-15T13:32:56.159123"} |
| {"step": 194, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.14, "mean_correctness_reward": 0.14, "mean_length_penalty": -0.00615234375, "mean_total_reward": 0.138154296875, "std_total_reward": 0.34702259133643976, "total_tokens": 630, "mean_completion_length": 12.6, "median_completion_length": 13, "min_completion_length": 10, "max_completion_length": 15, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0916390153486158, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.4001400470733643, "avg_step_time_sec": 13.100065488176247, "gpu_allocated_gb": 32.362863104, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2546.153755426407, "eta_sec": 78.60039292905748, "timestamp": "2026-02-15T13:32:57.553855"} |
| {"step": 195, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.0, "mean_correctness_reward": 0.0, "mean_length_penalty": -0.0057421875, "mean_total_reward": -0.00172265625, "std_total_reward": 0.00043608081061733485, "total_tokens": 588, "mean_completion_length": 11.76, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 30, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.0002004449726327369, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 1.0, "step_time_sec": 1.371023416519165, "avg_step_time_sec": 13.039916554475441, "gpu_allocated_gb": 32.362799104, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2547.5328979492188, "eta_sec": 65.19958277237721, "timestamp": "2026-02-15T13:32:58.974703"} |
| {"step": 196, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.2, "mean_correctness_reward": 0.2, "mean_length_penalty": -0.005439453125, "mean_total_reward": 0.1983681640625, "std_total_reward": 0.4000836944150709, "total_tokens": 557, "mean_completion_length": 11.14, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 29, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 0.00015338250132080185, "frac_prompts_all_correct": 0.2, "frac_prompts_all_incorrect": 0.8, "step_time_sec": 1.4203829765319824, "avg_step_time_sec": 12.980633219894099, "gpu_allocated_gb": 32.3628032, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2548.9588582515717, "eta_sec": 51.922532879576394, "timestamp": "2026-02-15T13:33:00.420225"} |
| {"step": 197, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.005244140625, "mean_total_reward": 0.39842675781250003, "std_total_reward": 0.48986686000207996, "total_tokens": 537, "mean_completion_length": 10.74, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 14, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 3.4788111583118904e-05, "frac_prompts_all_correct": 0.4, "frac_prompts_all_incorrect": 0.6, "step_time_sec": 1.4439430236816406, "avg_step_time_sec": 12.922071340725507, "gpu_allocated_gb": 32.362820608, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2550.409537792206, "eta_sec": 38.76621402217652, "timestamp": "2026-02-15T13:33:01.948433"} |
| {"step": 198, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.4, "mean_correctness_reward": 0.4, "mean_length_penalty": -0.005859375, "mean_total_reward": 0.3982421875, "std_total_reward": 0.489658778552736, "total_tokens": 600, "mean_completion_length": 12, "median_completion_length": 11, "min_completion_length": 10, "max_completion_length": 16, "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": 2.1256747245788574, "avg_step_time_sec": 12.867544085088403, "gpu_allocated_gb": 32.362978304, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2552.540811061859, "eta_sec": 25.735088170176805, "timestamp": "2026-02-15T13:33:03.919671"} |
| {"step": 199, "num_prompts": 5, "num_rollouts": 50, "accuracy": 0.0, "mean_correctness_reward": 0.0, "mean_length_penalty": -0.0058984375, "mean_total_reward": -0.0017695312499999998, "std_total_reward": 0.0004776738450513004, "total_tokens": 604, "mean_completion_length": 12.08, "median_completion_length": 10, "min_completion_length": 10, "max_completion_length": 19, "frac_has_think": 1.0, "frac_has_boxed": 1.0, "mean_reward_std_per_prompt": 1.435247896162022e-05, "frac_prompts_all_correct": 0.0, "frac_prompts_all_incorrect": 1.0, "step_time_sec": 1.3479735851287842, "avg_step_time_sec": 12.809656796143882, "gpu_allocated_gb": 32.362787328, "gpu_reserved_gb": 79.958114304, "gpu_peak_gb": 74.53288704, "elapsed_sec": 2553.894907951355, "eta_sec": 12.809656796143882, "timestamp": "2026-02-15T13:33:05.355386"} |
|
|
