{ "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": 1898.0, "completions/max_terminated_length": 1898.0, "completions/mean_length": 518.9375, "completions/mean_terminated_length": 518.9375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.0025, "frac_reward_zero_std": 0.0, "grad_norm": 3.6585705280303955, "learning_rate": 0.0, "loss": 0.0, "num_tokens": 133956.0, "reward": 0.2867187261581421, "reward_std": 0.257229208946228, "rewards/format_reward/mean": 0.3671875, "rewards/format_reward/std": 0.22257846593856812, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "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": 1489.0, "completions/max_terminated_length": 1489.0, "completions/mean_length": 391.234375, "completions/mean_terminated_length": 391.234375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.005, "frac_reward_zero_std": 0.125, "grad_norm": 10.50838851928711, "learning_rate": 2e-07, "loss": 0.0, "num_tokens": 217499.0, "reward": 0.30937498807907104, "reward_std": 0.27397435903549194, "rewards/format_reward/mean": 0.4375, "rewards/format_reward/std": 0.3021090030670166, "rewards/mcq_exact_match_reward/mean": 0.265625, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "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": 0.046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1614.0, "completions/mean_length": 588.796875, "completions/mean_terminated_length": 517.0327758789062, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0075, "frac_reward_zero_std": 0.0, "grad_norm": 9.119047164916992, "learning_rate": 4e-07, "loss": 0.0, "num_tokens": 343006.0, "reward": 0.34687498211860657, "reward_std": 0.44322875142097473, "rewards/format_reward/mean": 0.34375, "rewards/format_reward/std": 0.279951810836792, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1788.0, "completions/mean_length": 490.484375, "completions/mean_terminated_length": 440.2419128417969, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.01, "frac_reward_zero_std": 0.0, "grad_norm": 5.886940002441406, "learning_rate": 6e-07, "loss": 0.0, "num_tokens": 474205.0, "reward": 0.32109373807907104, "reward_std": 0.44282007217407227, "rewards/format_reward/mean": 0.3984375, "rewards/format_reward/std": 0.28423789143562317, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 1464.0, "completions/max_terminated_length": 1464.0, "completions/mean_length": 462.734375, "completions/mean_terminated_length": 462.734375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0125, "frac_reward_zero_std": 0.125, "grad_norm": 3.3544857501983643, "learning_rate": 8e-07, "loss": -0.0, "num_tokens": 582700.0, "reward": 0.23671874403953552, "reward_std": 0.3424985408782959, "rewards/format_reward/mean": 0.3359375, "rewards/format_reward/std": 0.2366211861371994, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "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": 0.0625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1944.0, "completions/mean_length": 586.90625, "completions/mean_terminated_length": 489.5000305175781, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.015, "frac_reward_zero_std": 0.0, "grad_norm": 6.5141072273254395, "learning_rate": 1e-06, "loss": 0.0, "num_tokens": 725414.0, "reward": 0.14765623211860657, "reward_std": 0.22827254235744476, "rewards/format_reward/mean": 0.2265625, "rewards/format_reward/std": 0.2807259261608124, "rewards/mcq_exact_match_reward/mean": 0.125, "rewards/mcq_exact_match_reward/std": 0.3333333432674408, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1483.0, "completions/mean_length": 528.09375, "completions/mean_terminated_length": 503.9682922363281, "completions/min_length": 94.0, "completions/min_terminated_length": 94.0, "epoch": 0.0175, "frac_reward_zero_std": 0.0, "grad_norm": 1.296814203262329, "learning_rate": 9.99726628670463e-07, "loss": -0.0, "num_tokens": 847164.0, "reward": 0.23749998211860657, "reward_std": 0.3825153112411499, "rewards/format_reward/mean": 0.34375, "rewards/format_reward/std": 0.233588308095932, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "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": 0.0, "completions/max_length": 1698.0, "completions/max_terminated_length": 1698.0, "completions/mean_length": 518.890625, "completions/mean_terminated_length": 518.890625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.02, "frac_reward_zero_std": 0.0, "grad_norm": 5.310608386993408, "learning_rate": 9.989068136093872e-07, "loss": 0.0, "num_tokens": 976797.0, "reward": 0.22734375298023224, "reward_std": 0.3119032382965088, "rewards/format_reward/mean": 0.3984375, "rewards/format_reward/std": 0.2978719174861908, "rewards/mcq_exact_match_reward/mean": 0.1875, "rewards/mcq_exact_match_reward/std": 0.39339789748191833, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1257.0, "completions/mean_length": 564.8125, "completions/mean_terminated_length": 541.2698974609375, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.0225, "frac_reward_zero_std": 0.0, "grad_norm": 6.675235748291016, "learning_rate": 9.975414512725056e-07, "loss": -0.0, "num_tokens": 1115497.0, "reward": 0.19453123211860657, "reward_std": 0.3237643837928772, "rewards/format_reward/mean": 0.3828125, "rewards/format_reward/std": 0.2634054720401764, "rewards/mcq_exact_match_reward/mean": 0.15625, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1669.0, "completions/mean_length": 605.890625, "completions/mean_terminated_length": 559.3709716796875, "completions/min_length": 7.0, "completions/min_terminated_length": 7.0, "epoch": 0.025, "frac_reward_zero_std": 0.0, "grad_norm": 2.0196316242218018, "learning_rate": 9.956320346634875e-07, "loss": -0.0, "num_tokens": 1248458.0, "reward": 0.25468748807907104, "reward_std": 0.4067697525024414, "rewards/format_reward/mean": 0.359375, "rewards/format_reward/std": 0.22658175230026245, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "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": 0.0, "completions/max_length": 1320.0, "completions/max_terminated_length": 1320.0, "completions/mean_length": 487.40625, "completions/mean_terminated_length": 487.40625, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.0275, "frac_reward_zero_std": 0.125, "grad_norm": 14.439066886901855, "learning_rate": 9.931806517013612e-07, "loss": -0.0, "num_tokens": 1364828.0, "reward": 0.24140623211860657, "reward_std": 0.2646661400794983, "rewards/format_reward/mean": 0.3828125, "rewards/format_reward/std": 0.21347814798355103, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "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": 0.046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1873.0, "completions/mean_length": 646.375, "completions/mean_terminated_length": 577.4425659179688, "completions/min_length": 10.0, "completions/min_terminated_length": 10.0, "epoch": 0.03, "frac_reward_zero_std": 0.0, "grad_norm": 2.276933193206787, "learning_rate": 9.901899829374047e-07, "loss": -0.0, "num_tokens": 1514876.0, "reward": 0.265625, "reward_std": 0.4345517158508301, "rewards/format_reward/mean": 0.46875, "rewards/format_reward/std": 0.25, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1568.0, "completions/mean_length": 520.359375, "completions/mean_terminated_length": 496.11114501953125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0325, "frac_reward_zero_std": 0.0, "grad_norm": 7.413666725158691, "learning_rate": 9.866632986240029e-07, "loss": -0.0, "num_tokens": 1636875.0, "reward": 0.26093748211860657, "reward_std": 0.3563947081565857, "rewards/format_reward/mean": 0.421875, "rewards/format_reward/std": 0.2847827076911926, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1721.0, "completions/mean_length": 468.1875, "completions/mean_terminated_length": 443.11114501953125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.035, "frac_reward_zero_std": 0.125, "grad_norm": 7.13982629776001, "learning_rate": 9.826044551386742e-07, "loss": -0.0, "num_tokens": 1773463.0, "reward": 0.4867187440395355, "reward_std": 0.37628647685050964, "rewards/format_reward/mean": 0.4921875, "rewards/format_reward/std": 0.3020063638687134, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2018.0, "completions/mean_length": 545.4375, "completions/mean_terminated_length": 521.5873413085938, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0375, "frac_reward_zero_std": 0.125, "grad_norm": 8.857248306274414, "learning_rate": 9.780178907671788e-07, "loss": 0.0, "num_tokens": 1888331.0, "reward": 0.16874998807907104, "reward_std": 0.22962015867233276, "rewards/format_reward/mean": 0.4375, "rewards/format_reward/std": 0.36187344789505005, "rewards/mcq_exact_match_reward/mean": 0.125, "rewards/mcq_exact_match_reward/std": 0.3333333432674408, "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": 0.0, "completions/max_length": 1677.0, "completions/max_terminated_length": 1677.0, "completions/mean_length": 522.328125, "completions/mean_terminated_length": 522.328125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.04, "frac_reward_zero_std": 0.125, "grad_norm": 4.947120666503906, "learning_rate": 9.729086208503173e-07, "loss": 0.0, "num_tokens": 2009232.0, "reward": 0.24062499403953552, "reward_std": 0.37447136640548706, "rewards/format_reward/mean": 0.53125, "rewards/format_reward/std": 0.265398770570755, "rewards/mcq_exact_match_reward/mean": 0.1875, "rewards/mcq_exact_match_reward/std": 0.39339789748191833, "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": 1043.0, "completions/max_terminated_length": 1043.0, "completions/mean_length": 384.84375, "completions/mean_terminated_length": 384.84375, "completions/min_length": 7.0, "completions/min_terminated_length": 7.0, "epoch": 0.0425, "frac_reward_zero_std": 0.0, "grad_norm": 10.11476993560791, "learning_rate": 9.672822322997304e-07, "loss": -0.0, "num_tokens": 2106918.0, "reward": 0.42265623807907104, "reward_std": 0.402811735868454, "rewards/format_reward/mean": 0.4765625, "rewards/format_reward/std": 0.3824491798877716, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 1453.0, "completions/mean_length": 495.3125, "completions/mean_terminated_length": 470.66668701171875, "completions/min_length": 13.0, "completions/min_terminated_length": 13.0, "epoch": 0.045, "frac_reward_zero_std": 0.0, "grad_norm": 3.7139668464660645, "learning_rate": 9.611448774886923e-07, "loss": 0.0, "num_tokens": 2228818.0, "reward": 0.22812500596046448, "reward_std": 0.2865009307861328, "rewards/format_reward/mean": 0.5625, "rewards/format_reward/std": 0.3726780116558075, "rewards/mcq_exact_match_reward/mean": 0.171875, "rewards/mcq_exact_match_reward/std": 0.38025420904159546, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1489.0, "completions/mean_length": 482.359375, "completions/mean_terminated_length": 431.8548278808594, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.0475, "frac_reward_zero_std": 0.0, "grad_norm": 24.99872589111328, "learning_rate": 9.545032675245813e-07, "loss": -0.0, "num_tokens": 2354041.0, "reward": 0.32890623807907104, "reward_std": 0.38791507482528687, "rewards/format_reward/mean": 0.4765625, "rewards/format_reward/std": 0.3719285726547241, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 1281.0, "completions/max_terminated_length": 1281.0, "completions/mean_length": 474.140625, "completions/mean_terminated_length": 474.140625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.05, "frac_reward_zero_std": 0.125, "grad_norm": 4.42039155960083, "learning_rate": 9.473646649103817e-07, "loss": 0.0, "num_tokens": 2473554.0, "reward": 0.4234374761581421, "reward_std": 0.4008023738861084, "rewards/format_reward/mean": 0.640625, "rewards/format_reward/std": 0.28824523091316223, "rewards/mcq_exact_match_reward/mean": 0.359375, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "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": 1559.0, "completions/mean_length": 463.140625, "completions/mean_terminated_length": 437.9841613769531, "completions/min_length": 13.0, "completions/min_terminated_length": 13.0, "epoch": 0.0525, "frac_reward_zero_std": 0.0, "grad_norm": 12.267646789550781, "learning_rate": 9.397368756032444e-07, "loss": 0.0, "num_tokens": 2581563.0, "reward": 0.3070312440395355, "reward_std": 0.29351624846458435, "rewards/format_reward/mean": 0.5703125, "rewards/format_reward/std": 0.36587780714035034, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1872.0, "completions/mean_length": 382.15625, "completions/mean_terminated_length": 355.71429443359375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.055, "frac_reward_zero_std": 0.0, "grad_norm": 13.092376708984375, "learning_rate": 9.316282404787869e-07, "loss": 0.0, "num_tokens": 2695637.0, "reward": 0.5023437738418579, "reward_std": 0.37737423181533813, "rewards/format_reward/mean": 0.6484375, "rewards/format_reward/std": 0.3044600784778595, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1873.0, "completions/mean_length": 486.1875, "completions/mean_terminated_length": 461.3968505859375, "completions/min_length": 25.0, "completions/min_terminated_length": 25.0, "epoch": 0.0575, "frac_reward_zero_std": 0.0, "grad_norm": 2.140165090560913, "learning_rate": 9.230476262104676e-07, "loss": -0.0, "num_tokens": 2807545.0, "reward": 0.36250001192092896, "reward_std": 0.42309853434562683, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.37796446681022644, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1439.0, "completions/mean_length": 510.34375, "completions/mean_terminated_length": 460.7419128417969, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.06, "frac_reward_zero_std": 0.0, "grad_norm": 10.928065299987793, "learning_rate": 9.1400441557401e-07, "loss": -0.0, "num_tokens": 2954975.0, "reward": 0.21796874701976776, "reward_std": 0.2888752222061157, "rewards/format_reward/mean": 0.6171875, "rewards/format_reward/std": 0.3302498757839203, "rewards/mcq_exact_match_reward/mean": 0.15625, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "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": 0.0, "completions/max_length": 892.0, "completions/max_terminated_length": 892.0, "completions/mean_length": 294.75, "completions/mean_terminated_length": 294.75, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0625, "frac_reward_zero_std": 0.0, "grad_norm": 11.743786811828613, "learning_rate": 9.045084971874737e-07, "loss": 0.0, "num_tokens": 3051815.0, "reward": 0.3656249940395355, "reward_std": 0.4116372764110565, "rewards/format_reward/mean": 0.6875, "rewards/format_reward/std": 0.3149704039096832, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1990.0, "completions/mean_length": 566.3125, "completions/mean_terminated_length": 518.51611328125, "completions/min_length": 13.0, "completions/min_terminated_length": 13.0, "epoch": 0.065, "frac_reward_zero_std": 0.0, "grad_norm": 8.691550254821777, "learning_rate": 8.945702546981968e-07, "loss": 0.0, "num_tokens": 3169315.0, "reward": 0.4312499761581421, "reward_std": 0.32975146174430847, "rewards/format_reward/mean": 0.5625, "rewards/format_reward/std": 0.33923351764678955, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 1352.0, "completions/max_terminated_length": 1352.0, "completions/mean_length": 441.40625, "completions/mean_terminated_length": 441.40625, "completions/min_length": 10.0, "completions/min_terminated_length": 10.0, "epoch": 0.0675, "frac_reward_zero_std": 0.0, "grad_norm": 10.450814247131348, "learning_rate": 8.842005554284295e-07, "loss": 0.0, "num_tokens": 3286469.0, "reward": 0.3968750238418579, "reward_std": 0.3035643696784973, "rewards/format_reward/mean": 0.6875, "rewards/format_reward/std": 0.3149704039096832, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 979.0, "completions/max_terminated_length": 979.0, "completions/mean_length": 397.5625, "completions/mean_terminated_length": 397.5625, "completions/min_length": 13.0, "completions/min_terminated_length": 13.0, "epoch": 0.07, "frac_reward_zero_std": 0.0, "grad_norm": 9.67840576171875, "learning_rate": 8.734107384920769e-07, "loss": 0.0, "num_tokens": 3392377.0, "reward": 0.45390623807907104, "reward_std": 0.424520879983902, "rewards/format_reward/mean": 0.6328125, "rewards/format_reward/std": 0.34780704975128174, "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": 1483.0, "completions/max_terminated_length": 1483.0, "completions/mean_length": 167.359375, "completions/mean_terminated_length": 167.359375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0725, "frac_reward_zero_std": 0.375, "grad_norm": 6.014416217803955, "learning_rate": 8.622126023955445e-07, "loss": 0.0, "num_tokens": 3497472.0, "reward": 0.4281250238418579, "reward_std": 0.24410976469516754, "rewards/format_reward/mean": 0.6875, "rewards/format_reward/std": 0.3149704039096832, "rewards/mcq_exact_match_reward/mean": 0.359375, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1940.0, "completions/mean_length": 280.75, "completions/mean_terminated_length": 252.69842529296875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.075, "frac_reward_zero_std": 0.0, "grad_norm": 27.0042667388916, "learning_rate": 8.506183921362442e-07, "loss": 0.0, "num_tokens": 3603776.0, "reward": 0.49296876788139343, "reward_std": 0.36379802227020264, "rewards/format_reward/mean": 0.8671875, "rewards/format_reward/std": 0.2987033724784851, "rewards/mcq_exact_match_reward/mean": 0.40625, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1462.0, "completions/mean_length": 307.78125, "completions/mean_terminated_length": 280.15875244140625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0775, "frac_reward_zero_std": 0.0, "grad_norm": 13.420266151428223, "learning_rate": 8.386407858128706e-07, "loss": -0.0, "num_tokens": 3737690.0, "reward": 0.4937500059604645, "reward_std": 0.45362722873687744, "rewards/format_reward/mean": 0.71875, "rewards/format_reward/std": 0.3069951832294464, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1931.0, "completions/mean_length": 326.28125, "completions/mean_terminated_length": 270.7419128417969, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.08, "frac_reward_zero_std": 0.25, "grad_norm": 12.415510177612305, "learning_rate": 8.262928807620843e-07, "loss": -0.0, "num_tokens": 3851196.0, "reward": 0.20859375596046448, "reward_std": 0.1422944962978363, "rewards/format_reward/mean": 0.8359375, "rewards/format_reward/std": 0.2824873626232147, "rewards/mcq_exact_match_reward/mean": 0.125, "rewards/mcq_exact_match_reward/std": 0.3333333432674408, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1427.0, "completions/mean_length": 137.765625, "completions/mean_terminated_length": 107.44445037841797, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0825, "frac_reward_zero_std": 0.125, "grad_norm": 17.117557525634766, "learning_rate": 8.135881792367685e-07, "loss": 0.0, "num_tokens": 3960957.0, "reward": 0.6484375, "reward_std": 0.3690624535083771, "rewards/format_reward/mean": 0.859375, "rewards/format_reward/std": 0.2592533528804779, "rewards/mcq_exact_match_reward/mean": 0.5625, "rewards/mcq_exact_match_reward/std": 0.5, "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": 920.0, "completions/mean_length": 187.4375, "completions/mean_terminated_length": 157.90476989746094, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.085, "frac_reward_zero_std": 0.25, "grad_norm": 10.581389427185059, "learning_rate": 8.005405736415125e-07, "loss": 0.0, "num_tokens": 4099177.0, "reward": 0.29218751192092896, "reward_std": 0.2078787088394165, "rewards/format_reward/mean": 0.734375, "rewards/format_reward/std": 0.30820462107658386, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "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": 801.0, "completions/max_terminated_length": 801.0, "completions/mean_length": 97.109375, "completions/mean_terminated_length": 97.109375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0875, "frac_reward_zero_std": 0.25, "grad_norm": 11.948210716247559, "learning_rate": 7.871643313414718e-07, "loss": -0.0, "num_tokens": 4183464.0, "reward": 0.6070312261581421, "reward_std": 0.31987857818603516, "rewards/format_reward/mean": 0.9140625, "rewards/format_reward/std": 0.209963858127594, "rewards/mcq_exact_match_reward/mean": 0.515625, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "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": 1071.0, "completions/max_terminated_length": 1071.0, "completions/mean_length": 146.328125, "completions/mean_terminated_length": 146.328125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.09, "frac_reward_zero_std": 0.25, "grad_norm": 11.833842277526855, "learning_rate": 7.734740790612136e-07, "loss": 0.0, "num_tokens": 4310357.0, "reward": 0.3984375, "reward_std": 0.3374851644039154, "rewards/format_reward/mean": 0.703125, "rewards/format_reward/std": 0.2630521357059479, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 961.0, "completions/max_terminated_length": 961.0, "completions/mean_length": 113.984375, "completions/mean_terminated_length": 113.984375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0925, "frac_reward_zero_std": 0.25, "grad_norm": 12.071688652038574, "learning_rate": 7.594847868906076e-07, "loss": 0.0, "num_tokens": 4468996.0, "reward": 0.3335937559604645, "reward_std": 0.19316625595092773, "rewards/format_reward/mean": 0.8359375, "rewards/format_reward/std": 0.29620200395584106, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 904.0, "completions/mean_length": 178.578125, "completions/mean_terminated_length": 148.90476989746094, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.095, "frac_reward_zero_std": 0.5, "grad_norm": 10.871786117553711, "learning_rate": 7.452117519152541e-07, "loss": 0.0, "num_tokens": 4562169.0, "reward": 0.24687501788139343, "reward_std": 0.23144522309303284, "rewards/format_reward/mean": 0.75, "rewards/format_reward/std": 0.35634833574295044, "rewards/mcq_exact_match_reward/mean": 0.171875, "rewards/mcq_exact_match_reward/std": 0.38025420904159546, "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": 736.0, "completions/max_terminated_length": 736.0, "completions/mean_length": 81.828125, "completions/mean_terminated_length": 81.828125, "completions/min_length": 11.0, "completions/min_terminated_length": 11.0, "epoch": 0.0975, "frac_reward_zero_std": 0.5, "grad_norm": 6.641048431396484, "learning_rate": 7.306705814893439e-07, "loss": -0.0, "num_tokens": 4644790.0, "reward": 0.46875, "reward_std": 0.18450236320495605, "rewards/format_reward/mean": 0.9375, "rewards/format_reward/std": 0.18898223340511322, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 0.0, "completions/max_length": 415.0, "completions/max_terminated_length": 415.0, "completions/mean_length": 60.921875, "completions/mean_terminated_length": 60.921875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1, "frac_reward_zero_std": 0.375, "grad_norm": 10.598438262939453, "learning_rate": 7.158771761692464e-07, "loss": -0.0, "num_tokens": 4726353.0, "reward": 0.62890625, "reward_std": 0.22785264253616333, "rewards/format_reward/mean": 0.9765625, "rewards/format_reward/std": 0.10652101784944534, "rewards/mcq_exact_match_reward/mean": 0.53125, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "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": 434.0, "completions/max_terminated_length": 434.0, "completions/mean_length": 32.0625, "completions/mean_terminated_length": 32.0625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1025, "frac_reward_zero_std": 0.375, "grad_norm": 14.398859977722168, "learning_rate": 7.008477123264847e-07, "loss": -0.0, "num_tokens": 4832237.0, "reward": 0.31953126192092896, "reward_std": 0.23467989265918732, "rewards/format_reward/mean": 0.8515625, "rewards/format_reward/std": 0.2302463799715042, "rewards/mcq_exact_match_reward/mean": 0.234375, "rewards/mcq_exact_match_reward/std": 0.42695629596710205, "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": 183.0, "completions/max_terminated_length": 183.0, "completions/mean_length": 17.71875, "completions/mean_terminated_length": 17.71875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.105, "frac_reward_zero_std": 0.25, "grad_norm": 18.40520668029785, "learning_rate": 6.855986244591103e-07, "loss": -0.0, "num_tokens": 4925459.0, "reward": 0.6296875476837158, "reward_std": 0.30840662121772766, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.08768405020236969, "rewards/mcq_exact_match_reward/mean": 0.53125, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "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": 274.0, "completions/max_terminated_length": 274.0, "completions/mean_length": 32.421875, "completions/mean_terminated_length": 32.421875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1075, "frac_reward_zero_std": 0.5, "grad_norm": 13.782668113708496, "learning_rate": 6.701465872208216e-07, "loss": -0.0, "num_tokens": 5037694.0, "reward": 0.6734375357627869, "reward_std": 0.20134371519088745, "rewards/format_reward/mean": 0.953125, "rewards/format_reward/std": 0.14689241349697113, "rewards/mcq_exact_match_reward/mean": 0.578125, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 17.0, "completions/max_terminated_length": 17.0, "completions/mean_length": 12.078125, "completions/mean_terminated_length": 12.078125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.11, "frac_reward_zero_std": 0.375, "grad_norm": 9.072900772094727, "learning_rate": 6.545084971874736e-07, "loss": 0.0, "num_tokens": 5124131.0, "reward": 0.3343750238418579, "reward_std": 0.3098084330558777, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.234375, "rewards/mcq_exact_match_reward/std": 0.42695629596710205, "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": 146.0, "completions/max_terminated_length": 146.0, "completions/mean_length": 15.4375, "completions/mean_terminated_length": 15.4375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1125, "frac_reward_zero_std": 0.625, "grad_norm": 25.2934627532959, "learning_rate": 6.387014543809223e-07, "loss": 0.0, "num_tokens": 5211639.0, "reward": 0.948437511920929, "reward_std": 0.13440215587615967, "rewards/format_reward/mean": 0.890625, "rewards/format_reward/std": 0.2083333432674408, "rewards/mcq_exact_match_reward/mean": 0.859375, "rewards/mcq_exact_match_reward/std": 0.3503824472427368, "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": 17.0, "completions/max_terminated_length": 17.0, "completions/mean_length": 12.234375, "completions/mean_terminated_length": 12.234375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.115, "frac_reward_zero_std": 0.375, "grad_norm": 17.072263717651367, "learning_rate": 6.227427435703995e-07, "loss": 0.0, "num_tokens": 5325310.0, "reward": 0.3968750238418579, "reward_std": 0.3098084330558777, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "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": 235.0, "completions/max_terminated_length": 235.0, "completions/mean_length": 19.75, "completions/mean_terminated_length": 19.75, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1175, "frac_reward_zero_std": 0.5, "grad_norm": 14.8613920211792, "learning_rate": 6.066498153718734e-07, "loss": 0.0, "num_tokens": 5401062.0, "reward": 0.7250000238418579, "reward_std": 0.24511480331420898, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.625, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 0.0, "completions/max_length": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.12, "frac_reward_zero_std": 0.625, "grad_norm": 7.733453750610352, "learning_rate": 5.90440267166055e-07, "loss": 0.0, "num_tokens": 5536718.0, "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": 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": 117.0, "completions/max_terminated_length": 117.0, "completions/mean_length": 13.03125, "completions/mean_terminated_length": 13.03125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1225, "frac_reward_zero_std": 0.75, "grad_norm": 7.888546466827393, "learning_rate": 5.741318238559209e-07, "loss": 0.0, "num_tokens": 5652632.0, "reward": 0.6937500238418579, "reward_std": 0.04857338219881058, "rewards/format_reward/mean": 0.84375, "rewards/format_reward/std": 0.279951810836792, "rewards/mcq_exact_match_reward/mean": 0.609375, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 272.0, "completions/max_terminated_length": 272.0, "completions/mean_length": 19.421875, "completions/mean_terminated_length": 19.421875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.125, "frac_reward_zero_std": 0.625, "grad_norm": 11.70719051361084, "learning_rate": 5.577423184847931e-07, "loss": 0.0, "num_tokens": 5746731.0, "reward": 0.4046875238418579, "reward_std": 0.11178100109100342, "rewards/format_reward/mean": 0.921875, "rewards/format_reward/std": 0.18298126757144928, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "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": 112.0, "completions/max_terminated_length": 112.0, "completions/mean_length": 13.5625, "completions/mean_terminated_length": 13.5625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1275, "frac_reward_zero_std": 0.875, "grad_norm": 7.1207122802734375, "learning_rate": 5.412896727361662e-07, "loss": 0.0, "num_tokens": 5842079.0, "reward": 0.6937500238418579, "reward_std": 0.0578637570142746, "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": 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": 0.0, "completions/max_length": 13.0, "completions/max_terminated_length": 13.0, "completions/mean_length": 12.015625, "completions/mean_terminated_length": 12.015625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.13, "frac_reward_zero_std": 0.625, "grad_norm": 11.432031631469727, "learning_rate": 5.247918773366111e-07, "loss": 0.0, "num_tokens": 5951392.0, "reward": 0.5687500238418579, "reward_std": 0.1735912710428238, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.46875, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "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": 13.0, "completions/max_terminated_length": 13.0, "completions/mean_length": 12.015625, "completions/mean_terminated_length": 12.015625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1325, "frac_reward_zero_std": 0.5, "grad_norm": 7.075219631195068, "learning_rate": 5.082669723831793e-07, "loss": 0.0, "num_tokens": 6047625.0, "reward": 0.3968750238418579, "reward_std": 0.19044628739356995, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "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": 165.0, "completions/max_terminated_length": 165.0, "completions/mean_length": 14.390625, "completions/mean_terminated_length": 14.390625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.135, "frac_reward_zero_std": 0.75, "grad_norm": 10.1735200881958, "learning_rate": 4.917330276168208e-07, "loss": 0.0, "num_tokens": 6138338.0, "reward": 0.7250000238418579, "reward_std": 0.1157275140285492, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.625, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 0.0, "completions/max_length": 110.0, "completions/max_terminated_length": 110.0, "completions/mean_length": 13.53125, "completions/mean_terminated_length": 13.53125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1375, "frac_reward_zero_std": 0.875, "grad_norm": 5.011144638061523, "learning_rate": 4.752081226633888e-07, "loss": 0.0, "num_tokens": 6225964.0, "reward": 0.6468750238418579, "reward_std": 0.0646936446428299, "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": 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": 17.0, "completions/max_terminated_length": 17.0, "completions/mean_length": 12.078125, "completions/mean_terminated_length": 12.078125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.14, "frac_reward_zero_std": 0.625, "grad_norm": 9.928509712219238, "learning_rate": 4.5871032726383385e-07, "loss": 0.0, "num_tokens": 6313329.0, "reward": 0.7093750238418579, "reward_std": 0.1530819982290268, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.609375, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 0.0, "completions/max_length": 17.0, "completions/max_terminated_length": 17.0, "completions/mean_length": 12.15625, "completions/mean_terminated_length": 12.15625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1425, "frac_reward_zero_std": 0.875, "grad_norm": 8.250155448913574, "learning_rate": 4.4225768151520694e-07, "loss": 0.0, "num_tokens": 6408675.0, "reward": 0.5062500238418579, "reward_std": 0.0578637570142746, "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": 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": 17.0, "completions/max_terminated_length": 17.0, "completions/mean_length": 12.15625, "completions/mean_terminated_length": 12.15625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.145, "frac_reward_zero_std": 0.5, "grad_norm": 13.277327537536621, "learning_rate": 4.258681761440789e-07, "loss": 0.0, "num_tokens": 6501253.0, "reward": 0.5531250238418579, "reward_std": 0.19044628739356995, "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": 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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1475, "frac_reward_zero_std": 0.625, "grad_norm": 9.02729606628418, "learning_rate": 4.095597328339452e-07, "loss": 0.0, "num_tokens": 6600861.0, "reward": 0.3656250238418579, "reward_std": 0.1530819833278656, "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": 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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.15, "frac_reward_zero_std": 0.375, "grad_norm": 17.841041564941406, "learning_rate": 3.9335018462812664e-07, "loss": 0.0, "num_tokens": 6671693.0, "reward": 0.4281250238418579, "reward_std": 0.28930896520614624, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 11.625, "completions/mean_terminated_length": 11.625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1525, "frac_reward_zero_std": 0.625, "grad_norm": 25.52366828918457, "learning_rate": 3.772572564296004e-07, "loss": -0.0, "num_tokens": 6788589.0, "reward": 0.34687501192092896, "reward_std": 0.17832404375076294, "rewards/format_reward/mean": 0.96875, "rewards/format_reward/std": 0.12198751419782639, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "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": 60.0, "completions/max_terminated_length": 60.0, "completions/mean_length": 13.21875, "completions/mean_terminated_length": 13.21875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.155, "frac_reward_zero_std": 0.375, "grad_norm": 21.121408462524414, "learning_rate": 3.612985456190778e-07, "loss": 0.0, "num_tokens": 6893971.0, "reward": 0.5476562976837158, "reward_std": 0.2802106738090515, "rewards/format_reward/mean": 0.9453125, "rewards/format_reward/std": 0.20152568817138672, "rewards/mcq_exact_match_reward/mean": 0.453125, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1575, "frac_reward_zero_std": 0.75, "grad_norm": 7.652548313140869, "learning_rate": 3.454915028125263e-07, "loss": 0.0, "num_tokens": 7028875.0, "reward": 0.6312500238418579, "reward_std": 0.10888782143592834, "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": 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": 38.0, "completions/max_terminated_length": 38.0, "completions/mean_length": 14.46875, "completions/mean_terminated_length": 14.46875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.16, "frac_reward_zero_std": 0.5, "grad_norm": 12.726404190063477, "learning_rate": 3.2985341277917846e-07, "loss": 0.0, "num_tokens": 7122401.0, "reward": 0.47968751192092896, "reward_std": 0.13700193166732788, "rewards/format_reward/mean": 0.890625, "rewards/format_reward/std": 0.3145764470100403, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 95.0, "completions/max_terminated_length": 95.0, "completions/mean_length": 13.375, "completions/mean_terminated_length": 13.375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1625, "frac_reward_zero_std": 0.75, "grad_norm": 6.691098213195801, "learning_rate": 3.1440137554088953e-07, "loss": -0.0, "num_tokens": 7200537.0, "reward": 0.3812500238418579, "reward_std": 0.0883883461356163, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.165, "frac_reward_zero_std": 0.875, "grad_norm": 5.075128555297852, "learning_rate": 2.9915228767351535e-07, "loss": -0.0, "num_tokens": 7305769.0, "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": 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": 0.0, "completions/max_length": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 11.25, "completions/mean_terminated_length": 11.25, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1675, "frac_reward_zero_std": 0.75, "grad_norm": 8.189491271972656, "learning_rate": 2.841228238307536e-07, "loss": 0.0, "num_tokens": 7389777.0, "reward": 0.8531250357627869, "reward_std": 0.10205793380737305, "rewards/format_reward/mean": 0.875, "rewards/format_reward/std": 0.3333333432674408, "rewards/mcq_exact_match_reward/mean": 0.765625, "rewards/mcq_exact_match_reward/std": 0.42695629596710205, "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": 14.0, "completions/max_terminated_length": 14.0, "completions/mean_length": 12.0625, "completions/mean_terminated_length": 12.0625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.17, "frac_reward_zero_std": 0.75, "grad_norm": 6.761085033416748, "learning_rate": 2.6932941851065615e-07, "loss": 0.0, "num_tokens": 7481301.0, "reward": 0.36250001192092896, "reward_std": 0.04998054727911949, "rewards/format_reward/mean": 0.96875, "rewards/format_reward/std": 0.17536810040473938, "rewards/mcq_exact_match_reward/mean": 0.265625, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "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": 67.0, "completions/max_terminated_length": 67.0, "completions/mean_length": 12.859375, "completions/mean_terminated_length": 12.859375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1725, "frac_reward_zero_std": 0.75, "grad_norm": 12.42138957977295, "learning_rate": 2.547882480847461e-07, "loss": -0.0, "num_tokens": 7571772.0, "reward": 0.7875000238418579, "reward_std": 0.10888782143592834, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.6875, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.175, "frac_reward_zero_std": 0.5, "grad_norm": 9.032736778259277, "learning_rate": 2.4051521310939254e-07, "loss": 0.0, "num_tokens": 7678764.0, "reward": 0.8968750238418579, "reward_std": 0.19044627249240875, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.796875, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "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": 248.0, "completions/max_terminated_length": 248.0, "completions/mean_length": 15.59375, "completions/mean_terminated_length": 15.59375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1775, "frac_reward_zero_std": 0.75, "grad_norm": 13.351987838745117, "learning_rate": 2.2652592093878665e-07, "loss": 0.0, "num_tokens": 7775738.0, "reward": 0.7546875476837158, "reward_std": 0.10676109790802002, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.08768405020236969, "rewards/mcq_exact_match_reward/mean": 0.65625, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 0.0, "completions/max_length": 13.0, "completions/max_terminated_length": 13.0, "completions/mean_length": 12.015625, "completions/mean_terminated_length": 12.015625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.18, "frac_reward_zero_std": 0.875, "grad_norm": 10.324533462524414, "learning_rate": 2.128356686585282e-07, "loss": -0.0, "num_tokens": 7863891.0, "reward": 0.6937500238418579, "reward_std": 0.0578637570142746, "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": 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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1825, "frac_reward_zero_std": 0.875, "grad_norm": 3.831482172012329, "learning_rate": 1.9945942635848745e-07, "loss": 0.0, "num_tokens": 7942419.0, "reward": 0.7093750238418579, "reward_std": 0.04419417306780815, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.609375, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.185, "frac_reward_zero_std": 0.375, "grad_norm": 16.41383171081543, "learning_rate": 1.8641182076323148e-07, "loss": 0.0, "num_tokens": 8055211.0, "reward": 0.5531250238418579, "reward_std": 0.29826053977012634, "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": 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": 17.0, "completions/max_terminated_length": 17.0, "completions/mean_length": 12.078125, "completions/mean_terminated_length": 12.078125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1875, "frac_reward_zero_std": 0.625, "grad_norm": 16.041906356811523, "learning_rate": 1.7370711923791564e-07, "loss": 0.0, "num_tokens": 8143104.0, "reward": 0.6296875476837158, "reward_std": 0.18868009746074677, "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": 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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.19, "frac_reward_zero_std": 0.75, "grad_norm": 10.163311958312988, "learning_rate": 1.6135921418712955e-07, "loss": 0.0, "num_tokens": 8209560.0, "reward": 0.46562501788139343, "reward_std": 0.09237537533044815, "rewards/format_reward/mean": 0.90625, "rewards/format_reward/std": 0.29378482699394226, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 11.8125, "completions/mean_terminated_length": 11.8125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1925, "frac_reward_zero_std": 0.5, "grad_norm": 13.508012771606445, "learning_rate": 1.493816078637557e-07, "loss": 0.0, "num_tokens": 8317916.0, "reward": 0.7078125476837158, "reward_std": 0.21688640117645264, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.08768405020236969, "rewards/mcq_exact_match_reward/mean": 0.609375, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "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": 8404852.0, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1975, "frac_reward_zero_std": 0.75, "grad_norm": 8.310019493103027, "learning_rate": 1.2658926150792322e-07, "loss": 0.0, "num_tokens": 8497092.0, "reward": 0.7406250238418579, "reward_std": 0.10205793380737305, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.640625, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2, "frac_reward_zero_std": 0.625, "grad_norm": 14.939309120178223, "learning_rate": 1.1579944457157059e-07, "loss": 0.0, "num_tokens": 8600372.0, "reward": 0.5375000238418579, "reward_std": 0.1825428307056427, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "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": 13.0, "completions/max_terminated_length": 13.0, "completions/mean_length": 12.015625, "completions/mean_terminated_length": 12.015625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2025, "frac_reward_zero_std": 0.375, "grad_norm": 15.373379707336426, "learning_rate": 1.0542974530180327e-07, "loss": 0.0, "num_tokens": 8674133.0, "reward": 0.48906251788139343, "reward_std": 0.2616810202598572, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.125, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.205, "frac_reward_zero_std": 0.75, "grad_norm": 24.713603973388672, "learning_rate": 9.549150281252632e-08, "loss": 0.0, "num_tokens": 8767397.0, "reward": 0.6937500238418579, "reward_std": 0.10888782143592834, "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": 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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2075, "frac_reward_zero_std": 0.75, "grad_norm": 14.414304733276367, "learning_rate": 8.599558442598998e-08, "loss": 0.0, "num_tokens": 8844189.0, "reward": 0.6468750238418579, "reward_std": 0.10205793380737305, "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": 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": 17.0, "completions/max_terminated_length": 17.0, "completions/mean_length": 12.078125, "completions/mean_terminated_length": 12.078125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.21, "frac_reward_zero_std": 0.75, "grad_norm": 8.169931411743164, "learning_rate": 7.695237378953224e-08, "loss": 0.0, "num_tokens": 8932930.0, "reward": 0.7250000238418579, "reward_std": 0.1157275140285492, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.625, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2125, "frac_reward_zero_std": 0.75, "grad_norm": 10.828977584838867, "learning_rate": 6.837175952121304e-08, "loss": 0.0, "num_tokens": 9005018.0, "reward": 0.3968750238418579, "reward_std": 0.11100947856903076, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "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": 9078986.0, "reward": 0.7250000238418579, "reward_std": 0.0, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.625, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 16.0, "completions/max_terminated_length": 16.0, "completions/mean_length": 12.125, "completions/mean_terminated_length": 12.125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2175, "frac_reward_zero_std": 0.75, "grad_norm": 8.94983196258545, "learning_rate": 5.263533508961826e-08, "loss": 0.0, "num_tokens": 9168850.0, "reward": 0.2828125059604645, "reward_std": 0.07328467071056366, "rewards/format_reward/mean": 0.953125, "rewards/format_reward/std": 0.21304203569889069, "rewards/mcq_exact_match_reward/mean": 0.1875, "rewards/mcq_exact_match_reward/std": 0.39339789748191833, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.22, "frac_reward_zero_std": 0.75, "grad_norm": 13.153602600097656, "learning_rate": 4.549673247541874e-08, "loss": -0.0, "num_tokens": 9262290.0, "reward": 0.6312500238418579, "reward_std": 0.0883883461356163, "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": 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": 168.0, "completions/max_terminated_length": 168.0, "completions/mean_length": 14.4375, "completions/mean_terminated_length": 14.4375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2225, "frac_reward_zero_std": 0.875, "grad_norm": 21.1671199798584, "learning_rate": 3.8855122511307626e-08, "loss": -0.0, "num_tokens": 9368638.0, "reward": 0.8656250238418579, "reward_std": 0.04419417306780815, "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": 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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.225, "frac_reward_zero_std": 0.625, "grad_norm": 15.989910125732422, "learning_rate": 3.271776770026963e-08, "loss": -0.0, "num_tokens": 9460670.0, "reward": 0.5218750238418579, "reward_std": 0.16887325048446655, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2275, "frac_reward_zero_std": 0.875, "grad_norm": 8.254137992858887, "learning_rate": 2.7091379149682682e-08, "loss": 0.0, "num_tokens": 9539110.0, "reward": 0.7562500238418579, "reward_std": 0.0578637570142746, "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": 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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.23, "frac_reward_zero_std": 0.5, "grad_norm": 14.285768508911133, "learning_rate": 2.1982109232821176e-08, "loss": -0.0, "num_tokens": 9635598.0, "reward": 0.2875000238418579, "reward_std": 0.1767766922712326, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.1875, "rewards/mcq_exact_match_reward/std": 0.39339789748191833, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2325, "frac_reward_zero_std": 0.75, "grad_norm": 16.36868667602539, "learning_rate": 1.7395544861325718e-08, "loss": 0.0, "num_tokens": 9714798.0, "reward": 1.006250023841858, "reward_std": 0.1246790662407875, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.90625, "rewards/mcq_exact_match_reward/std": 0.29378482699394226, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.235, "frac_reward_zero_std": 1.0, "grad_norm": 0.0, "learning_rate": 1.3336701375997127e-08, "loss": 0.0, "num_tokens": 9820270.0, "reward": 0.6000000238418579, "reward_std": 0.0, "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": 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": 334.0, "completions/max_terminated_length": 334.0, "completions/mean_length": 16.9375, "completions/mean_terminated_length": 16.9375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.2375, "frac_reward_zero_std": 0.625, "grad_norm": 12.444475173950195, "learning_rate": 9.810017062595321e-09, "loss": 0.0, "num_tokens": 9943074.0, "reward": 0.55078125, "reward_std": 0.18207663297653198, "rewards/format_reward/mean": 0.9765625, "rewards/format_reward/std": 0.13886408507823944, "rewards/mcq_exact_match_reward/mean": 0.453125, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.24, "frac_reward_zero_std": 0.75, "grad_norm": 8.018427848815918, "learning_rate": 6.819348298638839e-09, "loss": 0.0, "num_tokens": 10019866.0, "reward": 0.7875000238418579, "reward_std": 0.10888782143592834, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.6875, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2425, "frac_reward_zero_std": 0.25, "grad_norm": 17.901201248168945, "learning_rate": 4.367965336512403e-09, "loss": 0.0, "num_tokens": 10124898.0, "reward": 0.6156250238418579, "reward_std": 0.3266732692718506, "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": 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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.245, "frac_reward_zero_std": 0.5, "grad_norm": 11.791084289550781, "learning_rate": 2.458548727494292e-09, "loss": 0.0, "num_tokens": 10243674.0, "reward": 0.6625000238418579, "reward_std": 0.19727617502212524, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.2475, "frac_reward_zero_std": 0.375, "grad_norm": 11.973834991455078, "learning_rate": 1.0931863906127325e-09, "loss": 0.0, "num_tokens": 10345354.0, "reward": 0.3968750238418579, "reward_std": 0.24831004440784454, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.25, "frac_reward_zero_std": 0.75, "grad_norm": 12.83541488647461, "learning_rate": 2.733713295369755e-10, "loss": 0.0, "num_tokens": 10416842.0, "reward": 0.6937500238418579, "reward_std": 0.0883883461356163, "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": 100 } ], "logging_steps": 1, "max_steps": 100, "num_input_tokens_seen": 10416842, "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 }