{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 0.25, "eval_steps": 500, "global_step": 100, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1467.0, "completions/max_terminated_length": 1467.0, "completions/mean_length": 340.46875, "completions/mean_terminated_length": 340.46875, "completions/min_length": 2.0, "completions/min_terminated_length": 2.0, "epoch": 0.0025, "frac_reward_zero_std": 0.0, "grad_norm": 24.121065139770508, "learning_rate": 0.0, "loss": -0.0, "num_tokens": 104782.0, "reward": 0.2320312261581421, "reward_std": 0.2813003361225128, "rewards/format_reward/mean": 0.2890625, "rewards/format_reward/std": 0.2643453776836395, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "step": 1 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1311.0, "completions/max_terminated_length": 1311.0, "completions/mean_length": 349.15625, "completions/mean_terminated_length": 349.15625, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.005, "frac_reward_zero_std": 0.0, "grad_norm": 6.631368637084961, "learning_rate": 2e-07, "loss": -0.0, "num_tokens": 199528.0, "reward": 0.34218746423721313, "reward_std": 0.4091617465019226, "rewards/format_reward/mean": 0.296875, "rewards/format_reward/std": 0.2630521357059479, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "step": 2 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1933.0, "completions/max_terminated_length": 1933.0, "completions/mean_length": 378.484375, "completions/mean_terminated_length": 378.484375, "completions/min_length": 2.0, "completions/min_terminated_length": 2.0, "epoch": 0.0075, "frac_reward_zero_std": 0.0, "grad_norm": 22.637144088745117, "learning_rate": 4e-07, "loss": 0.0, "num_tokens": 342431.0, "reward": 0.43593746423721313, "reward_std": 0.44063660502433777, "rewards/format_reward/mean": 0.296875, "rewards/format_reward/std": 0.2777281701564789, "rewards/mcq_exact_match_reward/mean": 0.40625, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "step": 3 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1120.0, "completions/max_terminated_length": 1120.0, "completions/mean_length": 296.15625, "completions/mean_terminated_length": 296.15625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.01, "frac_reward_zero_std": 0.125, "grad_norm": 5.168887615203857, "learning_rate": 6e-07, "loss": 0.0, "num_tokens": 430457.0, "reward": 0.2671875059604645, "reward_std": 0.261038213968277, "rewards/format_reward/mean": 0.328125, "rewards/format_reward/std": 0.2847827076911926, "rewards/mcq_exact_match_reward/mean": 0.234375, "rewards/mcq_exact_match_reward/std": 0.42695629596710205, "step": 4 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1408.0, "completions/max_terminated_length": 1408.0, "completions/mean_length": 373.171875, "completions/mean_terminated_length": 373.171875, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.0125, "frac_reward_zero_std": 0.0, "grad_norm": 22.746238708496094, "learning_rate": 8e-07, "loss": -0.0, "num_tokens": 552004.0, "reward": 0.4624999761581421, "reward_std": 0.34055960178375244, "rewards/format_reward/mean": 0.40625, "rewards/format_reward/std": 0.233588308095932, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "step": 5 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2036.0, "completions/mean_length": 423.859375, "completions/mean_terminated_length": 398.0793762207031, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.015, "frac_reward_zero_std": 0.125, "grad_norm": 5.042137145996094, "learning_rate": 1e-06, "loss": -0.0, "num_tokens": 679451.0, "reward": 0.4781249761581421, "reward_std": 0.29430097341537476, "rewards/format_reward/mean": 0.40625, "rewards/format_reward/std": 0.233588308095932, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "step": 6 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1730.0, "completions/mean_length": 719.296875, "completions/mean_terminated_length": 630.7166748046875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0175, "frac_reward_zero_std": 0.0, "grad_norm": 2.657170295715332, "learning_rate": 9.99726628670463e-07, "loss": -0.0, "num_tokens": 842902.0, "reward": 0.25312498211860657, "reward_std": 0.3533875346183777, "rewards/format_reward/mean": 0.34375, "rewards/format_reward/std": 0.233588308095932, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "step": 7 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1097.0, "completions/mean_length": 528.5625, "completions/mean_terminated_length": 504.4444885253906, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.02, "frac_reward_zero_std": 0.0, "grad_norm": 7.923519611358643, "learning_rate": 9.989068136093872e-07, "loss": 0.0, "num_tokens": 967962.0, "reward": 0.24453124403953552, "reward_std": 0.24999795854091644, "rewards/format_reward/mean": 0.4140625, "rewards/format_reward/std": 0.2605654299259186, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "step": 8 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1769.0, "completions/max_terminated_length": 1769.0, "completions/mean_length": 430.078125, "completions/mean_terminated_length": 430.078125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0225, "frac_reward_zero_std": 0.0, "grad_norm": 9.529799461364746, "learning_rate": 9.975414512725056e-07, "loss": 0.0, "num_tokens": 1090047.0, "reward": 0.24296873807907104, "reward_std": 0.24107582867145538, "rewards/format_reward/mean": 0.3984375, "rewards/format_reward/std": 0.2387082874774933, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "step": 9 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1396.0, "completions/max_terminated_length": 1396.0, "completions/mean_length": 507.59375, "completions/mean_terminated_length": 507.59375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.025, "frac_reward_zero_std": 0.125, "grad_norm": 3.837048053741455, "learning_rate": 9.956320346634875e-07, "loss": -0.0, "num_tokens": 1221797.0, "reward": 0.5843749642372131, "reward_std": 0.3462211489677429, "rewards/format_reward/mean": 0.375, "rewards/format_reward/std": 0.2357022762298584, "rewards/mcq_exact_match_reward/mean": 0.546875, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "step": 10 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1354.0, "completions/mean_length": 652.03125, "completions/mean_terminated_length": 533.7288208007812, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0275, "frac_reward_zero_std": 0.0, "grad_norm": 3.759780168533325, "learning_rate": 9.931806517013612e-07, "loss": -0.0, "num_tokens": 1362423.0, "reward": 0.4781249463558197, "reward_std": 0.4550410211086273, "rewards/format_reward/mean": 0.40625, "rewards/format_reward/std": 0.25, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "step": 11 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1516.0, "completions/max_terminated_length": 1516.0, "completions/mean_length": 391.5625, "completions/mean_terminated_length": 391.5625, "completions/min_length": 5.0, "completions/min_terminated_length": 5.0, "epoch": 0.03, "frac_reward_zero_std": 0.0, "grad_norm": 7.199992656707764, "learning_rate": 9.901899829374047e-07, "loss": -0.0, "num_tokens": 1478563.0, "reward": 0.2851562201976776, "reward_std": 0.37806522846221924, "rewards/format_reward/mean": 0.3515625, "rewards/format_reward/std": 0.26246222853660583, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "step": 12 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1248.0, "completions/max_terminated_length": 1248.0, "completions/mean_length": 402.9375, "completions/mean_terminated_length": 402.9375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0325, "frac_reward_zero_std": 0.0, "grad_norm": 8.057401657104492, "learning_rate": 9.866632986240029e-07, "loss": 0.0, "num_tokens": 1587399.0, "reward": 0.3820312023162842, "reward_std": 0.36425960063934326, "rewards/format_reward/mean": 0.3828125, "rewards/format_reward/std": 0.2313210517168045, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 13 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1439.0, "completions/mean_length": 581.5, "completions/mean_terminated_length": 483.7333679199219, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.035, "frac_reward_zero_std": 0.0, "grad_norm": 1.4237709045410156, "learning_rate": 9.826044551386742e-07, "loss": -0.0, "num_tokens": 1717319.0, "reward": 0.4156249761581421, "reward_std": 0.37006694078445435, "rewards/format_reward/mean": 0.40625, "rewards/format_reward/std": 0.19669894874095917, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "step": 14 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1978.0, "completions/max_terminated_length": 1978.0, "completions/mean_length": 451.609375, "completions/mean_terminated_length": 451.609375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0375, "frac_reward_zero_std": 0.0, "grad_norm": 5.386332035064697, "learning_rate": 9.780178907671788e-07, "loss": -0.0, "num_tokens": 1822350.0, "reward": 0.23671874403953552, "reward_std": 0.3205496072769165, "rewards/format_reward/mean": 0.3359375, "rewards/format_reward/std": 0.2824873626232147, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "step": 15 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1391.0, "completions/mean_length": 554.8125, "completions/mean_terminated_length": 531.1111450195312, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.04, "frac_reward_zero_std": 0.0, "grad_norm": 6.627169609069824, "learning_rate": 9.729086208503173e-07, "loss": 0.0, "num_tokens": 1938426.0, "reward": 0.43437498807907104, "reward_std": 0.36538147926330566, "rewards/format_reward/mean": 0.4375, "rewards/format_reward/std": 0.1666666716337204, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "step": 16 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1754.0, "completions/max_terminated_length": 1754.0, "completions/mean_length": 504.859375, "completions/mean_terminated_length": 504.859375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0425, "frac_reward_zero_std": 0.125, "grad_norm": 10.752079963684082, "learning_rate": 9.672822322997304e-07, "loss": 0.0, "num_tokens": 2077889.0, "reward": 0.3359375, "reward_std": 0.29728609323501587, "rewards/format_reward/mean": 0.390625, "rewards/format_reward/std": 0.2083333432674408, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "step": 17 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1745.0, "completions/mean_length": 477.875, "completions/mean_terminated_length": 452.952392578125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.045, "frac_reward_zero_std": 0.125, "grad_norm": 15.702384948730469, "learning_rate": 9.611448774886923e-07, "loss": 0.0, "num_tokens": 2201481.0, "reward": 0.39296871423721313, "reward_std": 0.25133559107780457, "rewards/format_reward/mean": 0.4921875, "rewards/format_reward/std": 0.22699186205863953, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 18 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1560.0, "completions/mean_length": 479.328125, "completions/mean_terminated_length": 454.4285888671875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0475, "frac_reward_zero_std": 0.125, "grad_norm": 8.670381546020508, "learning_rate": 9.545032675245813e-07, "loss": 0.0, "num_tokens": 2328654.0, "reward": 0.26484373211860657, "reward_std": 0.2534114122390747, "rewards/format_reward/mean": 0.4609375, "rewards/format_reward/std": 0.18483558297157288, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "step": 19 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1680.0, "completions/max_terminated_length": 1680.0, "completions/mean_length": 339.71875, "completions/mean_terminated_length": 339.71875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.05, "frac_reward_zero_std": 0.125, "grad_norm": 7.443161487579346, "learning_rate": 9.473646649103817e-07, "loss": 0.0, "num_tokens": 2444524.0, "reward": 0.5249999761581421, "reward_std": 0.30251359939575195, "rewards/format_reward/mean": 0.5625, "rewards/format_reward/std": 0.22712838649749756, "rewards/mcq_exact_match_reward/mean": 0.46875, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "step": 20 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1120.0, "completions/mean_length": 241.296875, "completions/mean_terminated_length": 212.6190643310547, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0525, "frac_reward_zero_std": 0.5, "grad_norm": 11.614701271057129, "learning_rate": 9.397368756032444e-07, "loss": -0.0, "num_tokens": 2531023.0, "reward": 0.53125, "reward_std": 0.2136937975883484, "rewards/format_reward/mean": 0.46875, "rewards/format_reward/std": 0.17536810040473938, "rewards/mcq_exact_match_reward/mean": 0.484375, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 21 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1872.0, "completions/max_terminated_length": 1872.0, "completions/mean_length": 150.1875, "completions/mean_terminated_length": 150.1875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.055, "frac_reward_zero_std": 0.25, "grad_norm": 14.320113182067871, "learning_rate": 9.316282404787869e-07, "loss": 0.0, "num_tokens": 2643883.0, "reward": 0.53125, "reward_std": 0.3013129234313965, "rewards/format_reward/mean": 0.46875, "rewards/format_reward/std": 0.17536810040473938, "rewards/mcq_exact_match_reward/mean": 0.484375, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 22 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1496.0, "completions/max_terminated_length": 1496.0, "completions/mean_length": 89.8125, "completions/mean_terminated_length": 89.8125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0575, "frac_reward_zero_std": 0.5, "grad_norm": 15.463613510131836, "learning_rate": 9.230476262104676e-07, "loss": -0.0, "num_tokens": 2744167.0, "reward": 0.4718749523162842, "reward_std": 0.22673699259757996, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "step": 23 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 477.0, "completions/max_terminated_length": 477.0, "completions/mean_length": 32.15625, "completions/mean_terminated_length": 32.15625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.06, "frac_reward_zero_std": 0.625, "grad_norm": 17.539522171020508, "learning_rate": 9.1400441557401e-07, "loss": 0.0, "num_tokens": 2856569.0, "reward": 0.39296871423721313, "reward_std": 0.13159699738025665, "rewards/format_reward/mean": 0.4921875, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 24 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 881.0, "completions/mean_length": 150.96875, "completions/mean_terminated_length": 120.85714721679688, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0625, "frac_reward_zero_std": 0.125, "grad_norm": 22.964664459228516, "learning_rate": 9.045084971874737e-07, "loss": -0.0, "num_tokens": 2956095.0, "reward": 0.40703120827674866, "reward_std": 0.2988312840461731, "rewards/format_reward/mean": 0.4765625, "rewards/format_reward/std": 0.13886408507823944, "rewards/mcq_exact_match_reward/mean": 0.359375, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "step": 25 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 327.0, "completions/max_terminated_length": 327.0, "completions/mean_length": 18.0625, "completions/mean_terminated_length": 18.0625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.065, "frac_reward_zero_std": 0.5, "grad_norm": 21.533161163330078, "learning_rate": 8.945702546981968e-07, "loss": -0.0, "num_tokens": 3050171.0, "reward": 0.34687498211860657, "reward_std": 0.21778541803359985, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "step": 26 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1794.0, "completions/max_terminated_length": 1794.0, "completions/mean_length": 96.390625, "completions/mean_terminated_length": 96.390625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0675, "frac_reward_zero_std": 0.625, "grad_norm": 19.354248046875, "learning_rate": 8.842005554284295e-07, "loss": -0.0, "num_tokens": 3161332.0, "reward": 0.42499998211860657, "reward_std": 0.16675156354904175, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "step": 27 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 522.0, "completions/max_terminated_length": 522.0, "completions/mean_length": 36.171875, "completions/mean_terminated_length": 36.171875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.07, "frac_reward_zero_std": 0.75, "grad_norm": 4.9072980880737305, "learning_rate": 8.734107384920769e-07, "loss": 0.0, "num_tokens": 3233679.0, "reward": 0.43984371423721313, "reward_std": 0.046403877437114716, "rewards/format_reward/mean": 0.4921875, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "step": 28 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 438.0, "completions/max_terminated_length": 438.0, "completions/mean_length": 16.421875, "completions/mean_terminated_length": 16.421875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0725, "frac_reward_zero_std": 0.625, "grad_norm": 12.05967903137207, "learning_rate": 8.622126023955445e-07, "loss": -0.0, "num_tokens": 3352234.0, "reward": 0.25312498211860657, "reward_std": 0.15992167592048645, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "step": 29 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 957.0, "completions/max_terminated_length": 957.0, "completions/mean_length": 91.296875, "completions/mean_terminated_length": 91.296875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.075, "frac_reward_zero_std": 0.625, "grad_norm": 17.8170108795166, "learning_rate": 8.506183921362442e-07, "loss": -0.0, "num_tokens": 3459293.0, "reward": 0.5960937142372131, "reward_std": 0.1534513533115387, "rewards/format_reward/mean": 0.4921875, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.546875, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "step": 30 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1977.0, "completions/max_terminated_length": 1977.0, "completions/mean_length": 114.9375, "completions/mean_terminated_length": 114.9375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0775, "frac_reward_zero_std": 0.625, "grad_norm": 11.964653015136719, "learning_rate": 8.386407858128706e-07, "loss": -0.0, "num_tokens": 3566153.0, "reward": 0.4570312201976776, "reward_std": 0.09059804677963257, "rewards/format_reward/mean": 0.5078125, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.40625, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "step": 31 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 712.0, "completions/max_terminated_length": 712.0, "completions/mean_length": 69.9375, "completions/mean_terminated_length": 69.9375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.08, "frac_reward_zero_std": 0.625, "grad_norm": 13.64649772644043, "learning_rate": 8.262928807620843e-07, "loss": 0.0, "num_tokens": 3642269.0, "reward": 0.5523437261581421, "reward_std": 0.09162302315235138, "rewards/format_reward/mean": 0.5234375, "rewards/format_reward/std": 0.10652101784944534, "rewards/mcq_exact_match_reward/mean": 0.5, "rewards/mcq_exact_match_reward/std": 0.5039526224136353, "step": 32 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 593.0, "completions/max_terminated_length": 593.0, "completions/mean_length": 40.515625, "completions/mean_terminated_length": 40.515625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0825, "frac_reward_zero_std": 0.625, "grad_norm": 14.757148742675781, "learning_rate": 8.135881792367685e-07, "loss": -0.0, "num_tokens": 3736686.0, "reward": 0.4101562201976776, "reward_std": 0.10426762700080872, "rewards/format_reward/mean": 0.5078125, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.359375, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "step": 33 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 775.0, "completions/mean_length": 134.0, "completions/mean_terminated_length": 103.61905670166016, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.085, "frac_reward_zero_std": 0.5, "grad_norm": 12.825127601623535, "learning_rate": 8.005405736415125e-07, "loss": -0.0, "num_tokens": 3833502.0, "reward": 0.6906249523162842, "reward_std": 0.19939783215522766, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.640625, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "step": 34 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 609.0, "completions/max_terminated_length": 609.0, "completions/mean_length": 83.71875, "completions/mean_terminated_length": 83.71875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0875, "frac_reward_zero_std": 0.75, "grad_norm": 14.550989151000977, "learning_rate": 7.871643313414718e-07, "loss": -0.0, "num_tokens": 3934252.0, "reward": 0.6281249523162842, "reward_std": 0.10205792635679245, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.578125, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "step": 35 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 411.0, "completions/max_terminated_length": 411.0, "completions/mean_length": 52.0, "completions/mean_terminated_length": 52.0, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.09, "frac_reward_zero_std": 0.625, "grad_norm": 21.93817901611328, "learning_rate": 7.734740790612136e-07, "loss": 0.0, "num_tokens": 4037148.0, "reward": 0.5695312023162842, "reward_std": 0.13474363088607788, "rewards/format_reward/mean": 0.5390625, "rewards/format_reward/std": 0.13524486124515533, "rewards/mcq_exact_match_reward/mean": 0.515625, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 36 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 748.0, "completions/max_terminated_length": 748.0, "completions/mean_length": 105.453125, "completions/mean_terminated_length": 105.453125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0925, "frac_reward_zero_std": 0.5, "grad_norm": 13.75621509552002, "learning_rate": 7.594847868906076e-07, "loss": -0.0, "num_tokens": 4133001.0, "reward": 0.73828125, "reward_std": 0.23312518000602722, "rewards/format_reward/mean": 0.5078125, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.6875, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "step": 37 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 319.0, "completions/max_terminated_length": 319.0, "completions/mean_length": 39.96875, "completions/mean_terminated_length": 39.96875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.095, "frac_reward_zero_std": 0.5, "grad_norm": 10.660466194152832, "learning_rate": 7.452117519152541e-07, "loss": -0.0, "num_tokens": 4225415.0, "reward": 0.5929687023162842, "reward_std": 0.20632557570934296, "rewards/format_reward/mean": 0.6171875, "rewards/format_reward/std": 0.21347814798355103, "rewards/mcq_exact_match_reward/mean": 0.53125, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "step": 38 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 801.0, "completions/max_terminated_length": 801.0, "completions/mean_length": 112.671875, "completions/mean_terminated_length": 112.671875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0975, "frac_reward_zero_std": 0.75, "grad_norm": 11.935098648071289, "learning_rate": 7.306705814893439e-07, "loss": -0.0, "num_tokens": 4319626.0, "reward": 0.39296871423721313, "reward_std": 0.12520109117031097, "rewards/format_reward/mean": 0.4921875, "rewards/format_reward/std": 0.24384792149066925, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 39 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 667.0, "completions/mean_length": 128.875, "completions/mean_terminated_length": 98.41270446777344, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1, "frac_reward_zero_std": 0.625, "grad_norm": 12.141583442687988, "learning_rate": 7.158771761692464e-07, "loss": 0.0, "num_tokens": 4425530.0, "reward": 0.4468749761581421, "reward_std": 0.13258251547813416, "rewards/format_reward/mean": 0.5625, "rewards/format_reward/std": 0.1666666716337204, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "step": 40 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 618.0, "completions/max_terminated_length": 618.0, "completions/mean_length": 115.109375, "completions/mean_terminated_length": 115.109375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1025, "frac_reward_zero_std": 0.5, "grad_norm": 13.006111145019531, "learning_rate": 7.008477123264847e-07, "loss": 0.0, "num_tokens": 4515321.0, "reward": 0.37343746423721313, "reward_std": 0.16964474320411682, "rewards/format_reward/mean": 0.609375, "rewards/format_reward/std": 0.24346621334552765, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "step": 41 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 485.0, "completions/max_terminated_length": 485.0, "completions/mean_length": 78.375, "completions/mean_terminated_length": 78.375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.105, "frac_reward_zero_std": 0.375, "grad_norm": 32.34177017211914, "learning_rate": 6.855986244591103e-07, "loss": -0.0, "num_tokens": 4604545.0, "reward": 0.62890625, "reward_std": 0.23583920300006866, "rewards/format_reward/mean": 0.6640625, "rewards/format_reward/std": 0.2366211861371994, "rewards/mcq_exact_match_reward/mean": 0.5625, "rewards/mcq_exact_match_reward/std": 0.5, "step": 42 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 621.0, "completions/max_terminated_length": 621.0, "completions/mean_length": 113.390625, "completions/mean_terminated_length": 113.390625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1075, "frac_reward_zero_std": 0.75, "grad_norm": 13.46899127960205, "learning_rate": 6.701465872208216e-07, "loss": -0.0, "num_tokens": 4684994.0, "reward": 0.5203124284744263, "reward_std": 0.10834787786006927, "rewards/format_reward/mean": 0.515625, "rewards/format_reward/std": 0.08768405020236969, "rewards/mcq_exact_match_reward/mean": 0.46875, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "step": 43 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1031.0, "completions/max_terminated_length": 1031.0, "completions/mean_length": 148.828125, "completions/mean_terminated_length": 148.828125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.11, "frac_reward_zero_std": 0.75, "grad_norm": 16.012969970703125, "learning_rate": 6.545084971874736e-07, "loss": 0.0, "num_tokens": 4803671.0, "reward": 0.53125, "reward_std": 0.06154358386993408, "rewards/format_reward/mean": 0.625, "rewards/format_reward/std": 0.2357022762298584, "rewards/mcq_exact_match_reward/mean": 0.46875, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "step": 44 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 973.0, "completions/max_terminated_length": 973.0, "completions/mean_length": 105.0625, "completions/mean_terminated_length": 105.0625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1125, "frac_reward_zero_std": 0.125, "grad_norm": 24.89419937133789, "learning_rate": 6.387014543809223e-07, "loss": 0.0, "num_tokens": 4899051.0, "reward": 0.6148437261581421, "reward_std": 0.3771878480911255, "rewards/format_reward/mean": 0.6796875, "rewards/format_reward/std": 0.24180518090724945, "rewards/mcq_exact_match_reward/mean": 0.546875, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "step": 45 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 689.0, "completions/max_terminated_length": 689.0, "completions/mean_length": 112.140625, "completions/mean_terminated_length": 112.140625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.115, "frac_reward_zero_std": 0.25, "grad_norm": 53.93398666381836, "learning_rate": 6.227427435703995e-07, "loss": 0.0, "num_tokens": 4996884.0, "reward": 0.6117187738418579, "reward_std": 0.2741013467311859, "rewards/format_reward/mean": 0.6484375, "rewards/format_reward/std": 0.24688033759593964, "rewards/mcq_exact_match_reward/mean": 0.546875, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "step": 46 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 499.0, "completions/max_terminated_length": 499.0, "completions/mean_length": 41.1875, "completions/mean_terminated_length": 41.1875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1175, "frac_reward_zero_std": 0.5, "grad_norm": 22.220945358276367, "learning_rate": 6.066498153718734e-07, "loss": -0.0, "num_tokens": 5085840.0, "reward": 0.421875, "reward_std": 0.24641379714012146, "rewards/format_reward/mean": 0.78125, "rewards/format_reward/std": 0.25, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 47 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 815.0, "completions/mean_length": 232.984375, "completions/mean_terminated_length": 204.17462158203125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.12, "frac_reward_zero_std": 0.125, "grad_norm": 11.711411476135254, "learning_rate": 5.90440267166055e-07, "loss": 0.0, "num_tokens": 5208495.0, "reward": 0.34765625, "reward_std": 0.24685458838939667, "rewards/format_reward/mean": 0.6640625, "rewards/format_reward/std": 0.2680720090866089, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "step": 48 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 545.0, "completions/max_terminated_length": 545.0, "completions/mean_length": 65.984375, "completions/mean_terminated_length": 65.984375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1225, "frac_reward_zero_std": 0.5, "grad_norm": 21.655317306518555, "learning_rate": 5.741318238559209e-07, "loss": 0.0, "num_tokens": 5306174.0, "reward": 0.62890625, "reward_std": 0.11573586612939835, "rewards/format_reward/mean": 0.6640625, "rewards/format_reward/std": 0.2366211861371994, "rewards/mcq_exact_match_reward/mean": 0.5625, "rewards/mcq_exact_match_reward/std": 0.5, "step": 49 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 737.0, "completions/max_terminated_length": 737.0, "completions/mean_length": 79.53125, "completions/mean_terminated_length": 79.53125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.125, "frac_reward_zero_std": 0.375, "grad_norm": 14.4029541015625, "learning_rate": 5.577423184847931e-07, "loss": -0.0, "num_tokens": 5413376.0, "reward": 0.6429687142372131, "reward_std": 0.19145923852920532, "rewards/format_reward/mean": 0.8046875, "rewards/format_reward/std": 0.24587368965148926, "rewards/mcq_exact_match_reward/mean": 0.5625, "rewards/mcq_exact_match_reward/std": 0.5, "step": 50 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 408.0, "completions/max_terminated_length": 408.0, "completions/mean_length": 49.546875, "completions/mean_terminated_length": 49.546875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1275, "frac_reward_zero_std": 0.5, "grad_norm": 6.687479019165039, "learning_rate": 5.412896727361662e-07, "loss": -0.0, "num_tokens": 5515379.0, "reward": 0.7781250476837158, "reward_std": 0.15261505544185638, "rewards/format_reward/mean": 0.90625, "rewards/format_reward/std": 0.21593283116817474, "rewards/mcq_exact_match_reward/mean": 0.6875, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "step": 51 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 479.0, "completions/mean_length": 89.515625, "completions/mean_terminated_length": 58.42857360839844, "completions/min_length": 13.0, "completions/min_terminated_length": 13.0, "epoch": 0.13, "frac_reward_zero_std": 0.5, "grad_norm": 11.878891944885254, "learning_rate": 5.247918773366111e-07, "loss": -0.0, "num_tokens": 5596716.0, "reward": 0.36015626788139343, "reward_std": 0.05184826999902725, "rewards/format_reward/mean": 0.9453125, "rewards/format_reward/std": 0.15728822350502014, "rewards/mcq_exact_match_reward/mean": 0.265625, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "step": 52 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 647.0, "completions/max_terminated_length": 647.0, "completions/mean_length": 52.15625, "completions/mean_terminated_length": 52.15625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1325, "frac_reward_zero_std": 0.375, "grad_norm": 31.304433822631836, "learning_rate": 5.082669723831793e-07, "loss": 0.0, "num_tokens": 5693966.0, "reward": 0.680468738079071, "reward_std": 0.21469593048095703, "rewards/format_reward/mean": 0.8671875, "rewards/format_reward/std": 0.22257846593856812, "rewards/mcq_exact_match_reward/mean": 0.59375, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "step": 53 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1104.0, "completions/max_terminated_length": 1104.0, "completions/mean_length": 175.078125, "completions/mean_terminated_length": 175.078125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.135, "frac_reward_zero_std": 0.5, "grad_norm": 11.294580459594727, "learning_rate": 4.917330276168208e-07, "loss": 0.0, "num_tokens": 5816323.0, "reward": 0.40312501788139343, "reward_std": 0.12342280149459839, "rewards/format_reward/mean": 0.90625, "rewards/format_reward/std": 0.21593283116817474, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "step": 54 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 615.0, "completions/mean_length": 84.3125, "completions/mean_terminated_length": 53.142860412597656, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1375, "frac_reward_zero_std": 0.375, "grad_norm": 22.459693908691406, "learning_rate": 4.752081226633888e-07, "loss": 0.0, "num_tokens": 5906855.0, "reward": 0.9375, "reward_std": 0.208131343126297, "rewards/format_reward/mean": 0.9375, "rewards/format_reward/std": 0.18898223340511322, "rewards/mcq_exact_match_reward/mean": 0.84375, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "step": 55 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 577.0, "completions/max_terminated_length": 577.0, "completions/mean_length": 50.453125, "completions/mean_terminated_length": 50.453125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.14, "frac_reward_zero_std": 0.5, "grad_norm": 10.717191696166992, "learning_rate": 4.5871032726383385e-07, "loss": 0.0, "num_tokens": 5997684.0, "reward": 0.7156250476837158, "reward_std": 0.1756252646446228, "rewards/format_reward/mean": 0.90625, "rewards/format_reward/std": 0.19669894874095917, "rewards/mcq_exact_match_reward/mean": 0.625, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "step": 56 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 276.0, "completions/mean_length": 64.265625, "completions/mean_terminated_length": 32.77777862548828, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1425, "frac_reward_zero_std": 0.25, "grad_norm": 37.994075775146484, "learning_rate": 4.4225768151520694e-07, "loss": 0.0, "num_tokens": 6098893.0, "reward": 0.553906261920929, "reward_std": 0.26822715997695923, "rewards/format_reward/mean": 0.8515625, "rewards/format_reward/std": 0.24688033759593964, "rewards/mcq_exact_match_reward/mean": 0.46875, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "step": 57 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 739.0, "completions/max_terminated_length": 739.0, "completions/mean_length": 60.796875, "completions/mean_terminated_length": 60.796875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.145, "frac_reward_zero_std": 0.25, "grad_norm": 22.95911979675293, "learning_rate": 4.258681761440789e-07, "loss": 0.0, "num_tokens": 6184592.0, "reward": 0.6609375476837158, "reward_std": 0.27853259444236755, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.08768405020236969, "rewards/mcq_exact_match_reward/mean": 0.5625, "rewards/mcq_exact_match_reward/std": 0.5, "step": 58 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 479.0, "completions/max_terminated_length": 479.0, "completions/mean_length": 39.171875, "completions/mean_terminated_length": 39.171875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1475, "frac_reward_zero_std": 0.875, "grad_norm": 1.5430374145507812, "learning_rate": 4.095597328339452e-07, "loss": 0.0, "num_tokens": 6270011.0, "reward": 0.4593750238418579, "reward_std": 0.04419417306780815, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.359375, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "step": 59 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 205.0, "completions/max_terminated_length": 205.0, "completions/mean_length": 17.59375, "completions/mean_terminated_length": 17.59375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.15, "frac_reward_zero_std": 0.625, "grad_norm": 13.103647232055664, "learning_rate": 3.9335018462812664e-07, "loss": -0.0, "num_tokens": 6364745.0, "reward": 0.624218761920929, "reward_std": 0.09059804677963257, "rewards/format_reward/mean": 0.9296875, "rewards/format_reward/std": 0.1751912236213684, "rewards/mcq_exact_match_reward/mean": 0.53125, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "step": 60 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 775.0, "completions/max_terminated_length": 775.0, "completions/mean_length": 91.5, "completions/mean_terminated_length": 91.5, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1525, "frac_reward_zero_std": 0.25, "grad_norm": 15.906254768371582, "learning_rate": 3.772572564296004e-07, "loss": 0.0, "num_tokens": 6457257.0, "reward": 0.6000000238418579, "reward_std": 0.31300365924835205, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.5, "rewards/mcq_exact_match_reward/std": 0.5039526224136353, "step": 61 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 463.0, "completions/max_terminated_length": 463.0, "completions/mean_length": 32.828125, "completions/mean_terminated_length": 32.828125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.155, "frac_reward_zero_std": 0.375, "grad_norm": 12.228983879089355, "learning_rate": 3.612985456190778e-07, "loss": 0.0, "num_tokens": 6547750.0, "reward": 0.6453125476837158, "reward_std": 0.24789774417877197, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.125, "rewards/mcq_exact_match_reward/mean": 0.546875, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "step": 62 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 864.0, "completions/max_terminated_length": 864.0, "completions/mean_length": 51.96875, "completions/mean_terminated_length": 51.96875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1575, "frac_reward_zero_std": 0.625, "grad_norm": 10.3082857131958, "learning_rate": 3.454915028125263e-07, "loss": 0.0, "num_tokens": 6625652.0, "reward": 0.4437500238418579, "reward_std": 0.16675157845020294, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 63 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 404.0, "completions/max_terminated_length": 404.0, "completions/mean_length": 44.859375, "completions/mean_terminated_length": 44.859375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.16, "frac_reward_zero_std": 0.625, "grad_norm": 13.456597328186035, "learning_rate": 3.2985341277917846e-07, "loss": 0.0, "num_tokens": 6705731.0, "reward": 0.9046875238418579, "reward_std": 0.1976316273212433, "rewards/format_reward/mean": 0.921875, "rewards/format_reward/std": 0.20351573824882507, "rewards/mcq_exact_match_reward/mean": 0.8125, "rewards/mcq_exact_match_reward/std": 0.39339789748191833, "step": 64 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 138.0, "completions/max_terminated_length": 138.0, "completions/mean_length": 21.40625, "completions/mean_terminated_length": 21.40625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1625, "frac_reward_zero_std": 0.625, "grad_norm": 27.259490966796875, "learning_rate": 3.1440137554088953e-07, "loss": -0.0, "num_tokens": 6811901.0, "reward": 0.594531238079071, "reward_std": 0.08876051753759384, "rewards/format_reward/mean": 0.9453125, "rewards/format_reward/std": 0.15728822350502014, "rewards/mcq_exact_match_reward/mean": 0.5, "rewards/mcq_exact_match_reward/std": 0.5039526224136353, "step": 65 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 693.0, "completions/max_terminated_length": 693.0, "completions/mean_length": 42.546875, "completions/mean_terminated_length": 42.546875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.165, "frac_reward_zero_std": 0.625, "grad_norm": 40.02764129638672, "learning_rate": 2.9915228767351535e-07, "loss": -0.0, "num_tokens": 6911808.0, "reward": 0.6765625476837158, "reward_std": 0.06911306083202362, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.08768405020236969, "rewards/mcq_exact_match_reward/mean": 0.578125, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "step": 66 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 488.0, "completions/mean_length": 107.796875, "completions/mean_terminated_length": 45.20967483520508, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1675, "frac_reward_zero_std": 0.75, "grad_norm": 3.26263689994812, "learning_rate": 2.841228238307536e-07, "loss": 0.0, "num_tokens": 7015507.0, "reward": 0.5750000476837158, "reward_std": 0.04998054727911949, "rewards/format_reward/mean": 0.90625, "rewards/format_reward/std": 0.233588308095932, "rewards/mcq_exact_match_reward/mean": 0.484375, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 67 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 487.0, "completions/max_terminated_length": 487.0, "completions/mean_length": 40.65625, "completions/mean_terminated_length": 40.65625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.17, "frac_reward_zero_std": 0.625, "grad_norm": 12.225637435913086, "learning_rate": 2.6932941851065615e-07, "loss": 0.0, "num_tokens": 7129485.0, "reward": 0.7554687857627869, "reward_std": 0.14599090814590454, "rewards/format_reward/mean": 0.9921875, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.65625, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 68 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 349.0, "completions/max_terminated_length": 349.0, "completions/mean_length": 27.015625, "completions/mean_terminated_length": 27.015625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1725, "frac_reward_zero_std": 0.625, "grad_norm": 15.266570091247559, "learning_rate": 2.547882480847461e-07, "loss": 0.0, "num_tokens": 7228982.0, "reward": 0.8031250238418579, "reward_std": 0.13258251547813416, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.703125, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "step": 69 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 894.0, "completions/max_terminated_length": 894.0, "completions/mean_length": 34.828125, "completions/mean_terminated_length": 34.828125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.175, "frac_reward_zero_std": 0.5, "grad_norm": 19.81510353088379, "learning_rate": 2.4051521310939254e-07, "loss": 0.0, "num_tokens": 7323699.0, "reward": 0.7828124761581421, "reward_std": 0.17338219285011292, "rewards/format_reward/mean": 0.953125, "rewards/format_reward/std": 0.21304203569889069, "rewards/mcq_exact_match_reward/mean": 0.6875, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "step": 70 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 205.0, "completions/max_terminated_length": 205.0, "completions/mean_length": 26.625, "completions/mean_terminated_length": 26.625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1775, "frac_reward_zero_std": 0.625, "grad_norm": 18.10909080505371, "learning_rate": 2.2652592093878665e-07, "loss": 0.0, "num_tokens": 7437755.0, "reward": 0.6312500238418579, "reward_std": 0.16675157845020294, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.53125, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "step": 71 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1636.0, "completions/mean_length": 98.125, "completions/mean_terminated_length": 67.17460632324219, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.18, "frac_reward_zero_std": 0.375, "grad_norm": 18.8012752532959, "learning_rate": 2.128356686585282e-07, "loss": 0.0, "num_tokens": 7529235.0, "reward": 0.6296875476837158, "reward_std": 0.26496458053588867, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.125, "rewards/mcq_exact_match_reward/mean": 0.53125, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "step": 72 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 142.0, "completions/max_terminated_length": 142.0, "completions/mean_length": 15.140625, "completions/mean_terminated_length": 15.140625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1825, "frac_reward_zero_std": 0.625, "grad_norm": 12.054750442504883, "learning_rate": 1.9945942635848745e-07, "loss": 0.0, "num_tokens": 7627636.0, "reward": 0.3281250298023224, "reward_std": 0.1530819833278656, "rewards/format_reward/mean": 0.9375, "rewards/format_reward/std": 0.1666666716337204, "rewards/mcq_exact_match_reward/mean": 0.234375, "rewards/mcq_exact_match_reward/std": 0.42695629596710205, "step": 73 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 453.0, "completions/max_terminated_length": 453.0, "completions/mean_length": 42.171875, "completions/mean_terminated_length": 42.171875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.185, "frac_reward_zero_std": 0.75, "grad_norm": 5.343064785003662, "learning_rate": 1.8641182076323148e-07, "loss": -0.0, "num_tokens": 7713631.0, "reward": 0.5367187857627869, "reward_std": 0.10925719141960144, "rewards/format_reward/mean": 0.9921875, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "step": 74 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 405.0, "completions/max_terminated_length": 405.0, "completions/mean_length": 33.578125, "completions/mean_terminated_length": 33.578125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1875, "frac_reward_zero_std": 0.625, "grad_norm": 11.8765230178833, "learning_rate": 1.7370711923791564e-07, "loss": -0.0, "num_tokens": 7786860.0, "reward": 0.8656250238418579, "reward_std": 0.1530819833278656, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.765625, "rewards/mcq_exact_match_reward/std": 0.42695629596710205, "step": 75 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1352.0, "completions/max_terminated_length": 1352.0, "completions/mean_length": 67.3125, "completions/mean_terminated_length": 67.3125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.19, "frac_reward_zero_std": 0.5, "grad_norm": 13.052813529968262, "learning_rate": 1.6135921418712955e-07, "loss": -0.0, "num_tokens": 7875584.0, "reward": 0.518750011920929, "reward_std": 0.10750232636928558, "rewards/format_reward/mean": 0.96875, "rewards/format_reward/std": 0.12198751419782639, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "step": 76 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 306.0, "completions/max_terminated_length": 306.0, "completions/mean_length": 28.578125, "completions/mean_terminated_length": 28.578125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1925, "frac_reward_zero_std": 0.875, "grad_norm": 3.786604881286621, "learning_rate": 1.493816078637557e-07, "loss": -0.0, "num_tokens": 7959277.0, "reward": 0.6156250238418579, "reward_std": 0.04419417306780815, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.515625, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 77 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 357.0, "completions/max_terminated_length": 357.0, "completions/mean_length": 23.421875, "completions/mean_terminated_length": 23.421875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.195, "frac_reward_zero_std": 1.0, "grad_norm": 0.0, "learning_rate": 1.3778739760445552e-07, "loss": 0.0, "num_tokens": 8037752.0, "reward": 0.3500000238418579, "reward_std": 0.0, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "step": 78 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 391.0, "completions/max_terminated_length": 391.0, "completions/mean_length": 25.71875, "completions/mean_terminated_length": 25.71875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1975, "frac_reward_zero_std": 0.75, "grad_norm": 9.65241527557373, "learning_rate": 1.2658926150792322e-07, "loss": 0.0, "num_tokens": 8122046.0, "reward": 0.5531250238418579, "reward_std": 0.11100947856903076, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.453125, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "step": 79 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 311.0, "completions/max_terminated_length": 311.0, "completions/mean_length": 29.21875, "completions/mean_terminated_length": 29.21875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2, "frac_reward_zero_std": 0.875, "grad_norm": 2.3931398391723633, "learning_rate": 1.1579944457157059e-07, "loss": 0.0, "num_tokens": 8188148.0, "reward": 0.7234375476837158, "reward_std": 0.002893187804147601, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.08768405020236969, "rewards/mcq_exact_match_reward/mean": 0.625, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "step": 80 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 319.0, "completions/max_terminated_length": 319.0, "completions/mean_length": 32.984375, "completions/mean_terminated_length": 32.984375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2025, "frac_reward_zero_std": 0.625, "grad_norm": 14.807287216186523, "learning_rate": 1.0542974530180327e-07, "loss": 0.0, "num_tokens": 8271611.0, "reward": 0.6625000238418579, "reward_std": 0.16675157845020294, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.5625, "rewards/mcq_exact_match_reward/std": 0.5, "step": 81 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 104.0, "completions/max_terminated_length": 104.0, "completions/mean_length": 15.75, "completions/mean_terminated_length": 15.75, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.205, "frac_reward_zero_std": 0.75, "grad_norm": 11.109095573425293, "learning_rate": 9.549150281252632e-08, "loss": 0.0, "num_tokens": 8364723.0, "reward": 0.9125000238418579, "reward_std": 0.10888782143592834, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.8125, "rewards/mcq_exact_match_reward/std": 0.39339789748191833, "step": 82 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 187.0, "completions/max_terminated_length": 187.0, "completions/mean_length": 23.34375, "completions/mean_terminated_length": 23.34375, "completions/min_length": 11.0, "completions/min_terminated_length": 11.0, "epoch": 0.2075, "frac_reward_zero_std": 0.5, "grad_norm": 19.110769271850586, "learning_rate": 8.599558442598998e-08, "loss": 0.0, "num_tokens": 8458649.0, "reward": 0.9578125476837158, "reward_std": 0.21536517143249512, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.125, "rewards/mcq_exact_match_reward/mean": 0.859375, "rewards/mcq_exact_match_reward/std": 0.3503824472427368, "step": 83 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 563.0, "completions/max_terminated_length": 563.0, "completions/mean_length": 37.0, "completions/mean_terminated_length": 37.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.21, "frac_reward_zero_std": 0.375, "grad_norm": 17.616167068481445, "learning_rate": 7.695237378953224e-08, "loss": -0.0, "num_tokens": 8558449.0, "reward": 0.6117187738418579, "reward_std": 0.26363515853881836, "rewards/format_reward/mean": 0.9609375, "rewards/format_reward/std": 0.13524486124515533, "rewards/mcq_exact_match_reward/mean": 0.515625, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 84 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 299.0, "completions/max_terminated_length": 299.0, "completions/mean_length": 20.6875, "completions/mean_terminated_length": 20.6875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2125, "frac_reward_zero_std": 0.625, "grad_norm": 13.942946434020996, "learning_rate": 6.837175952121304e-08, "loss": 0.0, "num_tokens": 8628669.0, "reward": 0.5531250238418579, "reward_std": 0.1804211586713791, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.453125, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "step": 85 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 141.0, "completions/max_terminated_length": 141.0, "completions/mean_length": 14.265625, "completions/mean_terminated_length": 14.265625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.215, "frac_reward_zero_std": 1.0, "grad_norm": 0.0, "learning_rate": 6.026312439675551e-08, "loss": 0.0, "num_tokens": 8710598.0, "reward": 0.8500000238418579, "reward_std": 0.0, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.75, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "step": 86 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 203.0, "completions/max_terminated_length": 203.0, "completions/mean_length": 21.65625, "completions/mean_terminated_length": 21.65625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2175, "frac_reward_zero_std": 0.75, "grad_norm": 13.728775978088379, "learning_rate": 5.263533508961826e-08, "loss": 0.0, "num_tokens": 8809272.0, "reward": 0.5062500238418579, "reward_std": 0.0883883461356163, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.40625, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "step": 87 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 509.0, "completions/max_terminated_length": 509.0, "completions/mean_length": 49.875, "completions/mean_terminated_length": 49.875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.22, "frac_reward_zero_std": 0.75, "grad_norm": 5.599557876586914, "learning_rate": 4.549673247541874e-08, "loss": -0.0, "num_tokens": 8881728.0, "reward": 0.7515624761581421, "reward_std": 0.10764673352241516, "rewards/format_reward/mean": 0.953125, "rewards/format_reward/std": 0.14689241349697113, "rewards/mcq_exact_match_reward/mean": 0.65625, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 88 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 545.0, "completions/max_terminated_length": 545.0, "completions/mean_length": 33.375, "completions/mean_terminated_length": 33.375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2225, "frac_reward_zero_std": 0.75, "grad_norm": 6.32980489730835, "learning_rate": 3.8855122511307626e-08, "loss": 0.0, "num_tokens": 8968976.0, "reward": 0.6140625476837158, "reward_std": 0.04708736017346382, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.08768405020236969, "rewards/mcq_exact_match_reward/mean": 0.515625, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 89 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 289.0, "completions/max_terminated_length": 289.0, "completions/mean_length": 21.453125, "completions/mean_terminated_length": 21.453125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.225, "frac_reward_zero_std": 1.0, "grad_norm": 0.0, "learning_rate": 3.271776770026963e-08, "loss": 0.0, "num_tokens": 9047429.0, "reward": 0.8500000238418579, "reward_std": 0.0, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.75, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "step": 90 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 353.0, "completions/max_terminated_length": 353.0, "completions/mean_length": 24.234375, "completions/mean_terminated_length": 24.234375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2275, "frac_reward_zero_std": 0.625, "grad_norm": 11.335436820983887, "learning_rate": 2.7091379149682682e-08, "loss": -0.0, "num_tokens": 9138756.0, "reward": 0.5992187857627869, "reward_std": 0.09059805423021317, "rewards/format_reward/mean": 0.9921875, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.5, "rewards/mcq_exact_match_reward/std": 0.5039526224136353, "step": 91 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 218.0, "completions/max_terminated_length": 218.0, "completions/mean_length": 25.484375, "completions/mean_terminated_length": 25.484375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.23, "frac_reward_zero_std": 0.25, "grad_norm": 22.328004837036133, "learning_rate": 2.1982109232821176e-08, "loss": -0.0, "num_tokens": 9240619.0, "reward": 0.671875, "reward_std": 0.21536517143249512, "rewards/format_reward/mean": 0.9375, "rewards/format_reward/std": 0.1666666716337204, "rewards/mcq_exact_match_reward/mean": 0.578125, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "step": 92 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 205.0, "completions/max_terminated_length": 205.0, "completions/mean_length": 25.328125, "completions/mean_terminated_length": 25.328125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2325, "frac_reward_zero_std": 1.0, "grad_norm": 0.0, "learning_rate": 1.7395544861325718e-08, "loss": 0.0, "num_tokens": 9326800.0, "reward": 0.4750000238418579, "reward_std": 0.0, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "step": 93 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 499.0, "completions/max_terminated_length": 499.0, "completions/mean_length": 26.515625, "completions/mean_terminated_length": 26.515625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.235, "frac_reward_zero_std": 0.625, "grad_norm": 23.268037796020508, "learning_rate": 1.3336701375997127e-08, "loss": 0.0, "num_tokens": 9443873.0, "reward": 0.7562500238418579, "reward_std": 0.16675157845020294, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.65625, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 94 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 145.0, "completions/max_terminated_length": 145.0, "completions/mean_length": 14.328125, "completions/mean_terminated_length": 14.328125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2375, "frac_reward_zero_std": 0.25, "grad_norm": 15.48180103302002, "learning_rate": 9.810017062595321e-09, "loss": 0.0, "num_tokens": 9541870.0, "reward": 0.6156250238418579, "reward_std": 0.30617380142211914, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.515625, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 95 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 197.0, "completions/max_terminated_length": 197.0, "completions/mean_length": 27.671875, "completions/mean_terminated_length": 27.671875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.24, "frac_reward_zero_std": 0.875, "grad_norm": 2.1035268306732178, "learning_rate": 6.819348298638839e-09, "loss": -0.0, "num_tokens": 9636937.0, "reward": 0.3656250238418579, "reward_std": 0.04419417306780815, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.265625, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "step": 96 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 634.0, "completions/max_terminated_length": 634.0, "completions/mean_length": 49.171875, "completions/mean_terminated_length": 49.171875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2425, "frac_reward_zero_std": 0.375, "grad_norm": 16.285078048706055, "learning_rate": 4.367965336512403e-09, "loss": 0.0, "num_tokens": 9758708.0, "reward": 0.6468750238418579, "reward_std": 0.29826053977012634, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.546875, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "step": 97 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 311.0, "completions/max_terminated_length": 311.0, "completions/mean_length": 20.859375, "completions/mean_terminated_length": 20.859375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.245, "frac_reward_zero_std": 0.75, "grad_norm": 14.062127113342285, "learning_rate": 2.458548727494292e-09, "loss": 0.0, "num_tokens": 9854275.0, "reward": 0.6625000238418579, "reward_std": 0.1157275140285492, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.5625, "rewards/mcq_exact_match_reward/std": 0.5, "step": 98 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 354.0, "completions/max_terminated_length": 354.0, "completions/mean_length": 25.890625, "completions/mean_terminated_length": 25.890625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2475, "frac_reward_zero_std": 0.375, "grad_norm": 14.944952964782715, "learning_rate": 1.0931863906127325e-09, "loss": -0.0, "num_tokens": 9936580.0, "reward": 0.6937500238418579, "reward_std": 0.2346404641866684, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.59375, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "step": 99 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 17.0, "completions/max_terminated_length": 17.0, "completions/mean_length": 12.140625, "completions/mean_terminated_length": 12.140625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.25, "frac_reward_zero_std": 0.875, "grad_norm": 3.171335458755493, "learning_rate": 2.733713295369755e-10, "loss": -0.0, "num_tokens": 10024749.0, "reward": 0.8343750238418579, "reward_std": 0.04419417306780815, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.734375, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "step": 100 } ], "logging_steps": 1, "max_steps": 100, "num_input_tokens_seen": 10024749, "num_train_epochs": 1, "save_steps": 50, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": true }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 1, "trial_name": null, "trial_params": null }