{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 0.35486160397445, "eval_steps": 500, "global_step": 500, "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": 0.7734375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1999.0, "completions/mean_length": 1875.171875, "completions/mean_terminated_length": 1285.17236328125, "completions/min_length": 553.0, "completions/min_terminated_length": 553.0, "epoch": 0.0007097232079489, "frac_reward_zero_std": 0.4375, "grad_norm": 41.73765563964844, "learning_rate": 1e-06, "loss": 0.0526, "num_tokens": 2819342.0, "reward": 0.134765625, "reward_std": 0.21164482831954956, "rewards/simpleverify_reward/mean": 0.134765625, "rewards/simpleverify_reward/std": 0.3418070077896118, "step": 1, "tools/generated_tokens": 3383.796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.91796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.6484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2008.0, "completions/mean_length": 1733.427734375, "completions/mean_terminated_length": 1153.2166748046875, "completions/min_length": 255.0, "completions/min_terminated_length": 255.0, "epoch": 0.0014194464158978, "frac_reward_zero_std": 0.4375, "grad_norm": 75.10315704345703, "learning_rate": 9.98e-07, "loss": 0.0702, "num_tokens": 5407268.0, "reward": 0.26953125, "reward_std": 0.2468685805797577, "rewards/simpleverify_reward/mean": 0.26953125, "rewards/simpleverify_reward/std": 0.44415023922920227, "step": 2, "tools/generated_tokens": 2948.0390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.845703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.6171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1982.0, "completions/mean_length": 1725.3359375, "completions/mean_terminated_length": 1205.48974609375, "completions/min_length": 360.0, "completions/min_terminated_length": 360.0, "epoch": 0.0021291696238466998, "frac_reward_zero_std": 0.4375, "grad_norm": 46.71572494506836, "learning_rate": 9.959999999999999e-07, "loss": 0.0486, "num_tokens": 8002744.0, "reward": 0.236328125, "reward_std": 0.2245149463415146, "rewards/simpleverify_reward/mean": 0.236328125, "rewards/simpleverify_reward/std": 0.42524150013923645, "step": 3, "tools/generated_tokens": 3119.76953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.8125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.673828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1945.0, "completions/mean_length": 1784.32421875, "completions/mean_terminated_length": 1246.5509033203125, "completions/min_length": 480.0, "completions/min_terminated_length": 480.0, "epoch": 0.0028388928317956, "frac_reward_zero_std": 0.46875, "grad_norm": 33.60686111450195, "learning_rate": 9.94e-07, "loss": 0.0331, "num_tokens": 10644675.0, "reward": 0.1953125, "reward_std": 0.2148459255695343, "rewards/simpleverify_reward/mean": 0.1953125, "rewards/simpleverify_reward/std": 0.3968288004398346, "step": 4, "tools/generated_tokens": 3196.341796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.845703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.62109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1946.0, "completions/mean_length": 1704.685546875, "completions/mean_terminated_length": 1141.9329833984375, "completions/min_length": 346.0, "completions/min_terminated_length": 346.0, "epoch": 0.0035486160397444995, "frac_reward_zero_std": 0.46875, "grad_norm": 38.812625885009766, "learning_rate": 9.92e-07, "loss": 0.0413, "num_tokens": 13214122.0, "reward": 0.271484375, "reward_std": 0.22423109412193298, "rewards/simpleverify_reward/mean": 0.271484375, "rewards/simpleverify_reward/std": 0.44516023993492126, "step": 5, "tools/generated_tokens": 2993.6796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.818359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.646484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2003.0, "completions/mean_length": 1769.234375, "completions/mean_terminated_length": 1266.0220947265625, "completions/min_length": 343.0, "completions/min_terminated_length": 343.0, "epoch": 0.0042583392476933995, "frac_reward_zero_std": 0.34375, "grad_norm": 40.290531158447266, "learning_rate": 9.9e-07, "loss": 0.0371, "num_tokens": 15810787.0, "reward": 0.166015625, "reward_std": 0.22709707915782928, "rewards/simpleverify_reward/mean": 0.166015625, "rewards/simpleverify_reward/std": 0.3724585771560669, "step": 6, "tools/generated_tokens": 3062.578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.8359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.609375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1962.0, "completions/mean_length": 1649.232421875, "completions/mean_terminated_length": 1031.2550048828125, "completions/min_length": 172.0, "completions/min_terminated_length": 172.0, "epoch": 0.0049680624556423, "frac_reward_zero_std": 0.5, "grad_norm": 22.716703414916992, "learning_rate": 9.88e-07, "loss": 0.0406, "num_tokens": 18281259.0, "reward": 0.189453125, "reward_std": 0.20748455822467804, "rewards/simpleverify_reward/mean": 0.189453125, "rewards/simpleverify_reward/std": 0.3922513723373413, "step": 7, "tools/generated_tokens": 2898.361328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.783203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.681640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1914.0, "completions/mean_length": 1759.357421875, "completions/mean_terminated_length": 1141.343505859375, "completions/min_length": 458.0, "completions/min_terminated_length": 458.0, "epoch": 0.0056777856635912, "frac_reward_zero_std": 0.4375, "grad_norm": 44.68397903442383, "learning_rate": 9.86e-07, "loss": 0.0415, "num_tokens": 20931888.0, "reward": 0.24609375, "reward_std": 0.21521681547164917, "rewards/simpleverify_reward/mean": 0.24609375, "rewards/simpleverify_reward/std": 0.4311550557613373, "step": 8, "tools/generated_tokens": 3159.990234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.85546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.697265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1974.0, "completions/mean_length": 1797.04296875, "completions/mean_terminated_length": 1219.0322265625, "completions/min_length": 252.0, "completions/min_terminated_length": 252.0, "epoch": 0.0063875088715401, "frac_reward_zero_std": 0.40625, "grad_norm": 40.8090705871582, "learning_rate": 9.84e-07, "loss": 0.0776, "num_tokens": 23540825.0, "reward": 0.251953125, "reward_std": 0.2382301390171051, "rewards/simpleverify_reward/mean": 0.251953125, "rewards/simpleverify_reward/std": 0.43455907702445984, "step": 9, "tools/generated_tokens": 3072.26171875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.8359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.642578125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1918.0, "completions/mean_length": 1734.625, "completions/mean_terminated_length": 1180.218505859375, "completions/min_length": 263.0, "completions/min_terminated_length": 263.0, "epoch": 0.007097232079488999, "frac_reward_zero_std": 0.40625, "grad_norm": 48.90374755859375, "learning_rate": 9.819999999999999e-07, "loss": 0.0706, "num_tokens": 26162335.0, "reward": 0.1875, "reward_std": 0.2524182200431824, "rewards/simpleverify_reward/mean": 0.1875, "rewards/simpleverify_reward/std": 0.39069411158561707, "step": 10, "tools/generated_tokens": 3103.765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.849609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.59375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2008.0, "completions/mean_length": 1723.669921875, "completions/mean_terminated_length": 1249.6490478515625, "completions/min_length": 392.0, "completions/min_terminated_length": 392.0, "epoch": 0.007806955287437899, "frac_reward_zero_std": 0.3125, "grad_norm": 30.532155990600586, "learning_rate": 9.8e-07, "loss": 0.0623, "num_tokens": 28716841.0, "reward": 0.294921875, "reward_std": 0.2720392346382141, "rewards/simpleverify_reward/mean": 0.294921875, "rewards/simpleverify_reward/std": 0.4564536213874817, "step": 11, "tools/generated_tokens": 3042.837890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.798828125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.744140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1966.0, "completions/mean_length": 1843.806640625, "completions/mean_terminated_length": 1249.9312744140625, "completions/min_length": 242.0, "completions/min_terminated_length": 242.0, "epoch": 0.008516678495386799, "frac_reward_zero_std": 0.46875, "grad_norm": 35.94465637207031, "learning_rate": 9.78e-07, "loss": 0.0554, "num_tokens": 31485115.0, "reward": 0.193359375, "reward_std": 0.21139007806777954, "rewards/simpleverify_reward/mean": 0.193359375, "rewards/simpleverify_reward/std": 0.39531853795051575, "step": 12, "tools/generated_tokens": 3364.33984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.912109375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1979.0, "completions/mean_length": 1720.669921875, "completions/mean_terminated_length": 1137.16845703125, "completions/min_length": 282.0, "completions/min_terminated_length": 282.0, "epoch": 0.009226401703335699, "frac_reward_zero_std": 0.5, "grad_norm": 27.793285369873047, "learning_rate": 9.759999999999998e-07, "loss": 0.0387, "num_tokens": 34081377.0, "reward": 0.228515625, "reward_std": 0.1975235939025879, "rewards/simpleverify_reward/mean": 0.228515625, "rewards/simpleverify_reward/std": 0.4202871024608612, "step": 13, "tools/generated_tokens": 3058.361328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.83203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.517578125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1992.0, "completions/mean_length": 1629.0859375, "completions/mean_terminated_length": 1179.643798828125, "completions/min_length": 89.0, "completions/min_terminated_length": 89.0, "epoch": 0.0099361249112846, "frac_reward_zero_std": 0.21875, "grad_norm": 55.608150482177734, "learning_rate": 9.74e-07, "loss": 0.0985, "num_tokens": 36410489.0, "reward": 0.3203125, "reward_std": 0.297268807888031, "rewards/simpleverify_reward/mean": 0.3203125, "rewards/simpleverify_reward/std": 0.4670529365539551, "step": 14, "tools/generated_tokens": 2716.658203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.658203125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1943.0, "completions/mean_length": 1682.607421875, "completions/mean_terminated_length": 978.9656982421875, "completions/min_length": 209.0, "completions/min_terminated_length": 209.0, "epoch": 0.0106458481192335, "frac_reward_zero_std": 0.5, "grad_norm": 42.058937072753906, "learning_rate": 9.72e-07, "loss": 0.0399, "num_tokens": 38923668.0, "reward": 0.228515625, "reward_std": 0.19686490297317505, "rewards/simpleverify_reward/mean": 0.228515625, "rewards/simpleverify_reward/std": 0.4202871024608612, "step": 15, "tools/generated_tokens": 2969.3125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.806640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.55078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2002.0, "completions/mean_length": 1598.525390625, "completions/mean_terminated_length": 1047.430419921875, "completions/min_length": 382.0, "completions/min_terminated_length": 382.0, "epoch": 0.0113555713271824, "frac_reward_zero_std": 0.3125, "grad_norm": 41.21215057373047, "learning_rate": 9.7e-07, "loss": 0.0736, "num_tokens": 41349063.0, "reward": 0.3046875, "reward_std": 0.27545076608657837, "rewards/simpleverify_reward/mean": 0.3046875, "rewards/simpleverify_reward/std": 0.4607250988483429, "step": 16, "tools/generated_tokens": 2894.373046875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.765625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.755859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1986.0, "completions/mean_length": 1868.64453125, "completions/mean_terminated_length": 1313.360107421875, "completions/min_length": 295.0, "completions/min_terminated_length": 295.0, "epoch": 0.0120652945351313, "frac_reward_zero_std": 0.375, "grad_norm": 18.572778701782227, "learning_rate": 9.679999999999999e-07, "loss": 0.028, "num_tokens": 44186585.0, "reward": 0.15234375, "reward_std": 0.2390626072883606, "rewards/simpleverify_reward/mean": 0.15234375, "rewards/simpleverify_reward/std": 0.35970520973205566, "step": 17, "tools/generated_tokens": 3480.59375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.916015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1969.0, "completions/mean_length": 1751.19140625, "completions/mean_terminated_length": 1143.4405517578125, "completions/min_length": 295.0, "completions/min_terminated_length": 295.0, "epoch": 0.0127750177430802, "frac_reward_zero_std": 0.46875, "grad_norm": 51.31765365600586, "learning_rate": 9.66e-07, "loss": 0.0462, "num_tokens": 46826099.0, "reward": 0.205078125, "reward_std": 0.23933403193950653, "rewards/simpleverify_reward/mean": 0.205078125, "rewards/simpleverify_reward/std": 0.4041535556316376, "step": 18, "tools/generated_tokens": 3170.81640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.853515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.7109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1978.0, "completions/mean_length": 1794.02734375, "completions/mean_terminated_length": 1169.3919677734375, "completions/min_length": 262.0, "completions/min_terminated_length": 262.0, "epoch": 0.0134847409510291, "frac_reward_zero_std": 0.5, "grad_norm": 17.946596145629883, "learning_rate": 9.64e-07, "loss": 0.0303, "num_tokens": 49457801.0, "reward": 0.1484375, "reward_std": 0.17544928193092346, "rewards/simpleverify_reward/mean": 0.1484375, "rewards/simpleverify_reward/std": 0.35588082671165466, "step": 19, "tools/generated_tokens": 3205.326171875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.81640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.736328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1933.0, "completions/mean_length": 1831.04296875, "completions/mean_terminated_length": 1225.1702880859375, "completions/min_length": 286.0, "completions/min_terminated_length": 286.0, "epoch": 0.014194464158977998, "frac_reward_zero_std": 0.5, "grad_norm": 16.606714248657227, "learning_rate": 9.619999999999999e-07, "loss": 0.0217, "num_tokens": 52253999.0, "reward": 0.1171875, "reward_std": 0.1752775013446808, "rewards/simpleverify_reward/mean": 0.1171875, "rewards/simpleverify_reward/std": 0.32195815443992615, "step": 20, "tools/generated_tokens": 3420.9140625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.916015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1899.0, "completions/mean_length": 1713.3984375, "completions/mean_terminated_length": 1030.386962890625, "completions/min_length": 342.0, "completions/min_terminated_length": 342.0, "epoch": 0.014904187366926898, "frac_reward_zero_std": 0.4375, "grad_norm": 53.058006286621094, "learning_rate": 9.6e-07, "loss": 0.0611, "num_tokens": 54782214.0, "reward": 0.23046875, "reward_std": 0.21109169721603394, "rewards/simpleverify_reward/mean": 0.23046875, "rewards/simpleverify_reward/std": 0.42154473066329956, "step": 21, "tools/generated_tokens": 2967.45703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.8046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.5390625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1941.0, "completions/mean_length": 1617.08984375, "completions/mean_terminated_length": 1113.14404296875, "completions/min_length": 356.0, "completions/min_terminated_length": 356.0, "epoch": 0.015613910574875798, "frac_reward_zero_std": 0.28125, "grad_norm": 37.76161193847656, "learning_rate": 9.58e-07, "loss": 0.0862, "num_tokens": 57182462.0, "reward": 0.298828125, "reward_std": 0.2855975031852722, "rewards/simpleverify_reward/mean": 0.298828125, "rewards/simpleverify_reward/std": 0.45819199085235596, "step": 22, "tools/generated_tokens": 2865.626953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.720703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.62109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1984.0, "completions/mean_length": 1697.65234375, "completions/mean_terminated_length": 1125.597900390625, "completions/min_length": 240.0, "completions/min_terminated_length": 240.0, "epoch": 0.016323633782824698, "frac_reward_zero_std": 0.28125, "grad_norm": 35.11011505126953, "learning_rate": 9.559999999999998e-07, "loss": 0.066, "num_tokens": 59733257.0, "reward": 0.275390625, "reward_std": 0.28314200043678284, "rewards/simpleverify_reward/mean": 0.275390625, "rewards/simpleverify_reward/std": 0.44714778661727905, "step": 23, "tools/generated_tokens": 2997.68359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.833984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.6171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1980.0, "completions/mean_length": 1741.4140625, "completions/mean_terminated_length": 1247.1224365234375, "completions/min_length": 471.0, "completions/min_terminated_length": 471.0, "epoch": 0.017033356990773598, "frac_reward_zero_std": 0.375, "grad_norm": 38.042198181152344, "learning_rate": 9.539999999999999e-07, "loss": 0.0641, "num_tokens": 62294676.0, "reward": 0.287109375, "reward_std": 0.24907034635543823, "rewards/simpleverify_reward/mean": 0.287109375, "rewards/simpleverify_reward/std": 0.45285552740097046, "step": 24, "tools/generated_tokens": 3104.39453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.783203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.673828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1965.0, "completions/mean_length": 1741.564453125, "completions/mean_terminated_length": 1108.509033203125, "completions/min_length": 307.0, "completions/min_terminated_length": 307.0, "epoch": 0.017743080198722498, "frac_reward_zero_std": 0.59375, "grad_norm": 12.270814895629883, "learning_rate": 9.52e-07, "loss": 0.0235, "num_tokens": 64932490.0, "reward": 0.169921875, "reward_std": 0.16869445145130157, "rewards/simpleverify_reward/mean": 0.169921875, "rewards/simpleverify_reward/std": 0.3759314715862274, "step": 25, "tools/generated_tokens": 3197.98046875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.841796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.634765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1942.0, "completions/mean_length": 1733.4296875, "completions/mean_terminated_length": 1186.716552734375, "completions/min_length": 205.0, "completions/min_terminated_length": 205.0, "epoch": 0.018452803406671398, "frac_reward_zero_std": 0.46875, "grad_norm": 21.738588333129883, "learning_rate": 9.499999999999999e-07, "loss": 0.0284, "num_tokens": 67506579.0, "reward": 0.22265625, "reward_std": 0.21264970302581787, "rewards/simpleverify_reward/mean": 0.22265625, "rewards/simpleverify_reward/std": 0.41643625497817993, "step": 26, "tools/generated_tokens": 3109.921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.810546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.72265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1989.0, "completions/mean_length": 1796.10546875, "completions/mean_terminated_length": 1139.760498046875, "completions/min_length": 369.0, "completions/min_terminated_length": 369.0, "epoch": 0.0191625266146203, "frac_reward_zero_std": 0.40625, "grad_norm": 31.69468116760254, "learning_rate": 9.479999999999999e-07, "loss": 0.0388, "num_tokens": 70181012.0, "reward": 0.193359375, "reward_std": 0.23452921211719513, "rewards/simpleverify_reward/mean": 0.193359375, "rewards/simpleverify_reward/std": 0.39531853795051575, "step": 27, "tools/generated_tokens": 3253.55859375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.84375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.662109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1947.0, "completions/mean_length": 1707.619140625, "completions/mean_terminated_length": 1040.6300048828125, "completions/min_length": 252.0, "completions/min_terminated_length": 252.0, "epoch": 0.0198722498225692, "frac_reward_zero_std": 0.5, "grad_norm": 16.64727020263672, "learning_rate": 9.459999999999999e-07, "loss": 0.0366, "num_tokens": 72789747.0, "reward": 0.224609375, "reward_std": 0.20669373869895935, "rewards/simpleverify_reward/mean": 0.224609375, "rewards/simpleverify_reward/std": 0.41773295402526855, "step": 28, "tools/generated_tokens": 3163.099609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.822265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.53515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1984.0, "completions/mean_length": 1665.935546875, "completions/mean_terminated_length": 1228.3740234375, "completions/min_length": 213.0, "completions/min_terminated_length": 213.0, "epoch": 0.0205819730305181, "frac_reward_zero_std": 0.28125, "grad_norm": 29.882719039916992, "learning_rate": 9.439999999999999e-07, "loss": 0.0764, "num_tokens": 75349237.0, "reward": 0.396484375, "reward_std": 0.3220316767692566, "rewards/simpleverify_reward/mean": 0.396484375, "rewards/simpleverify_reward/std": 0.4896455705165863, "step": 29, "tools/generated_tokens": 3110.361328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.8125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.6875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1973.0, "completions/mean_length": 1800.5234375, "completions/mean_terminated_length": 1256.0750732421875, "completions/min_length": 469.0, "completions/min_terminated_length": 469.0, "epoch": 0.021291696238467, "frac_reward_zero_std": 0.40625, "grad_norm": 26.34286880493164, "learning_rate": 9.419999999999999e-07, "loss": 0.0387, "num_tokens": 78022607.0, "reward": 0.173828125, "reward_std": 0.21005403995513916, "rewards/simpleverify_reward/mean": 0.173828125, "rewards/simpleverify_reward/std": 0.3793322443962097, "step": 30, "tools/generated_tokens": 3239.02734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.84375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.6875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1945.0, "completions/mean_length": 1775.595703125, "completions/mean_terminated_length": 1176.3062744140625, "completions/min_length": 376.0, "completions/min_terminated_length": 376.0, "epoch": 0.0220014194464159, "frac_reward_zero_std": 0.375, "grad_norm": 25.385038375854492, "learning_rate": 9.399999999999999e-07, "loss": 0.0483, "num_tokens": 80670090.0, "reward": 0.2421875, "reward_std": 0.2460412085056305, "rewards/simpleverify_reward/mean": 0.2421875, "rewards/simpleverify_reward/std": 0.42882615327835083, "step": 31, "tools/generated_tokens": 3252.40625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.841796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.720703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1958.0, "completions/mean_length": 1855.490234375, "completions/mean_terminated_length": 1358.7342529296875, "completions/min_length": 455.0, "completions/min_terminated_length": 455.0, "epoch": 0.0227111426543648, "frac_reward_zero_std": 0.5625, "grad_norm": 9.318013191223145, "learning_rate": 9.379999999999998e-07, "loss": 0.0226, "num_tokens": 83454208.0, "reward": 0.17578125, "reward_std": 0.1791287660598755, "rewards/simpleverify_reward/mean": 0.17578125, "rewards/simpleverify_reward/std": 0.3810062110424042, "step": 32, "tools/generated_tokens": 3440.884765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.876953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.595703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1949.0, "completions/mean_length": 1690.74609375, "completions/mean_terminated_length": 1164.357421875, "completions/min_length": 276.0, "completions/min_terminated_length": 276.0, "epoch": 0.0234208658623137, "frac_reward_zero_std": 0.375, "grad_norm": 46.448909759521484, "learning_rate": 9.36e-07, "loss": 0.0428, "num_tokens": 86005672.0, "reward": 0.30078125, "reward_std": 0.24274347722530365, "rewards/simpleverify_reward/mean": 0.30078125, "rewards/simpleverify_reward/std": 0.45904624462127686, "step": 33, "tools/generated_tokens": 3062.623046875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.822265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.55859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1922.0, "completions/mean_length": 1645.056640625, "completions/mean_terminated_length": 1136.1282958984375, "completions/min_length": 378.0, "completions/min_terminated_length": 378.0, "epoch": 0.0241305890702626, "frac_reward_zero_std": 0.34375, "grad_norm": 27.743682861328125, "learning_rate": 9.34e-07, "loss": 0.0479, "num_tokens": 88530369.0, "reward": 0.28515625, "reward_std": 0.2762618958950043, "rewards/simpleverify_reward/mean": 0.28515625, "rewards/simpleverify_reward/std": 0.45193037390708923, "step": 34, "tools/generated_tokens": 2994.419921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.78515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.52734375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1977.0, "completions/mean_length": 1665.30078125, "completions/mean_terminated_length": 1242.6611328125, "completions/min_length": 265.0, "completions/min_terminated_length": 265.0, "epoch": 0.0248403122782115, "frac_reward_zero_std": 0.3125, "grad_norm": 14.042314529418945, "learning_rate": 9.32e-07, "loss": 0.0656, "num_tokens": 91074019.0, "reward": 0.32421875, "reward_std": 0.2910993695259094, "rewards/simpleverify_reward/mean": 0.32421875, "rewards/simpleverify_reward/std": 0.4685399830341339, "step": 35, "tools/generated_tokens": 3131.435546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.8046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.55859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1972.0, "completions/mean_length": 1675.822265625, "completions/mean_terminated_length": 1204.8363037109375, "completions/min_length": 326.0, "completions/min_terminated_length": 326.0, "epoch": 0.0255500354861604, "frac_reward_zero_std": 0.40625, "grad_norm": 37.203514099121094, "learning_rate": 9.3e-07, "loss": 0.0376, "num_tokens": 93569259.0, "reward": 0.296875, "reward_std": 0.24448150396347046, "rewards/simpleverify_reward/mean": 0.296875, "rewards/simpleverify_reward/std": 0.45732781291007996, "step": 36, "tools/generated_tokens": 2969.654296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.662109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1925.0, "completions/mean_length": 1743.70703125, "completions/mean_terminated_length": 1147.4334716796875, "completions/min_length": 432.0, "completions/min_terminated_length": 432.0, "epoch": 0.0262597586941093, "frac_reward_zero_std": 0.46875, "grad_norm": 9.813057899475098, "learning_rate": 9.28e-07, "loss": 0.033, "num_tokens": 96201935.0, "reward": 0.2265625, "reward_std": 0.215084046125412, "rewards/simpleverify_reward/mean": 0.2265625, "rewards/simpleverify_reward/std": 0.4190165400505066, "step": 37, "tools/generated_tokens": 3224.05859375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.830078125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.51171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1974.0, "completions/mean_length": 1586.41015625, "completions/mean_terminated_length": 1109.864013671875, "completions/min_length": 275.0, "completions/min_terminated_length": 275.0, "epoch": 0.0269694819020582, "frac_reward_zero_std": 0.375, "grad_norm": 21.908842086791992, "learning_rate": 9.26e-07, "loss": 0.0386, "num_tokens": 98554216.0, "reward": 0.349609375, "reward_std": 0.27093058824539185, "rewards/simpleverify_reward/mean": 0.349609375, "rewards/simpleverify_reward/std": 0.47731292247772217, "step": 38, "tools/generated_tokens": 2800.66796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7109375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.66015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1975.0, "completions/mean_length": 1754.53515625, "completions/mean_terminated_length": 1184.4713134765625, "completions/min_length": 221.0, "completions/min_terminated_length": 221.0, "epoch": 0.027679205110007096, "frac_reward_zero_std": 0.5, "grad_norm": 19.671567916870117, "learning_rate": 9.24e-07, "loss": 0.046, "num_tokens": 101222335.0, "reward": 0.244140625, "reward_std": 0.19200226664543152, "rewards/simpleverify_reward/mean": 0.244140625, "rewards/simpleverify_reward/std": 0.42999663949012756, "step": 39, "tools/generated_tokens": 3264.326171875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.841796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.662109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1984.0, "completions/mean_length": 1799.734375, "completions/mean_terminated_length": 1313.24853515625, "completions/min_length": 473.0, "completions/min_terminated_length": 473.0, "epoch": 0.028388928317955996, "frac_reward_zero_std": 0.40625, "grad_norm": 14.723760604858398, "learning_rate": 9.22e-07, "loss": 0.0452, "num_tokens": 103963417.0, "reward": 0.1953125, "reward_std": 0.2302786260843277, "rewards/simpleverify_reward/mean": 0.1953125, "rewards/simpleverify_reward/std": 0.3968288004398346, "step": 40, "tools/generated_tokens": 3367.419921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.888671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.6015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1975.0, "completions/mean_length": 1700.521484375, "completions/mean_terminated_length": 1175.8970947265625, "completions/min_length": 261.0, "completions/min_terminated_length": 261.0, "epoch": 0.029098651525904896, "frac_reward_zero_std": 0.4375, "grad_norm": 8.524008750915527, "learning_rate": 9.2e-07, "loss": 0.0288, "num_tokens": 106509735.0, "reward": 0.1796875, "reward_std": 0.2223920375108719, "rewards/simpleverify_reward/mean": 0.1796875, "rewards/simpleverify_reward/std": 0.38430243730545044, "step": 41, "tools/generated_tokens": 3089.388671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.80859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.609375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1996.0, "completions/mean_length": 1749.962890625, "completions/mean_terminated_length": 1285.0250244140625, "completions/min_length": 489.0, "completions/min_terminated_length": 489.0, "epoch": 0.029808374733853796, "frac_reward_zero_std": 0.375, "grad_norm": 19.685123443603516, "learning_rate": 9.18e-07, "loss": 0.0384, "num_tokens": 109107965.0, "reward": 0.234375, "reward_std": 0.2398073822259903, "rewards/simpleverify_reward/mean": 0.234375, "rewards/simpleverify_reward/std": 0.42402184009552, "step": 42, "tools/generated_tokens": 3169.431640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.6640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2006.0, "completions/mean_length": 1756.380859375, "completions/mean_terminated_length": 1179.9244384765625, "completions/min_length": 316.0, "completions/min_terminated_length": 316.0, "epoch": 0.030518097941802696, "frac_reward_zero_std": 0.375, "grad_norm": 11.85417366027832, "learning_rate": 9.16e-07, "loss": 0.0279, "num_tokens": 111798200.0, "reward": 0.234375, "reward_std": 0.21555635333061218, "rewards/simpleverify_reward/mean": 0.234375, "rewards/simpleverify_reward/std": 0.42402184009552, "step": 43, "tools/generated_tokens": 3282.845703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.873046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.697265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1992.0, "completions/mean_length": 1808.384765625, "completions/mean_terminated_length": 1256.4967041015625, "completions/min_length": 317.0, "completions/min_terminated_length": 317.0, "epoch": 0.031227821149751596, "frac_reward_zero_std": 0.5, "grad_norm": 6.051669597625732, "learning_rate": 9.14e-07, "loss": 0.0286, "num_tokens": 114492847.0, "reward": 0.193359375, "reward_std": 0.19971179962158203, "rewards/simpleverify_reward/mean": 0.193359375, "rewards/simpleverify_reward/std": 0.39531853795051575, "step": 44, "tools/generated_tokens": 3336.140625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.8515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.47265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1990.0, "completions/mean_length": 1610.05859375, "completions/mean_terminated_length": 1217.5333251953125, "completions/min_length": 473.0, "completions/min_terminated_length": 473.0, "epoch": 0.0319375443577005, "frac_reward_zero_std": 0.46875, "grad_norm": 7.054818153381348, "learning_rate": 9.12e-07, "loss": 0.0224, "num_tokens": 116803254.0, "reward": 0.337890625, "reward_std": 0.19327980279922485, "rewards/simpleverify_reward/mean": 0.337890625, "rewards/simpleverify_reward/std": 0.4734536409378052, "step": 45, "tools/generated_tokens": 2811.55078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.67578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.64453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1948.0, "completions/mean_length": 1774.751953125, "completions/mean_terminated_length": 1279.30224609375, "completions/min_length": 487.0, "completions/min_terminated_length": 487.0, "epoch": 0.032647267565649396, "frac_reward_zero_std": 0.375, "grad_norm": 11.977477073669434, "learning_rate": 9.1e-07, "loss": 0.0279, "num_tokens": 119468704.0, "reward": 0.255859375, "reward_std": 0.2371734380722046, "rewards/simpleverify_reward/mean": 0.255859375, "rewards/simpleverify_reward/std": 0.43676990270614624, "step": 46, "tools/generated_tokens": 3277.59375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.83203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.45703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1956.0, "completions/mean_length": 1599.751953125, "completions/mean_terminated_length": 1224.269775390625, "completions/min_length": 193.0, "completions/min_terminated_length": 193.0, "epoch": 0.0333569907735983, "frac_reward_zero_std": 0.3125, "grad_norm": 24.292966842651367, "learning_rate": 9.08e-07, "loss": 0.0627, "num_tokens": 121821664.0, "reward": 0.3671875, "reward_std": 0.27222147583961487, "rewards/simpleverify_reward/mean": 0.3671875, "rewards/simpleverify_reward/std": 0.48250964283943176, "step": 47, "tools/generated_tokens": 2849.478515625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.701171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.60546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1968.0, "completions/mean_length": 1718.451171875, "completions/mean_terminated_length": 1212.7078857421875, "completions/min_length": 335.0, "completions/min_terminated_length": 335.0, "epoch": 0.034066713981547196, "frac_reward_zero_std": 0.40625, "grad_norm": 41.17572021484375, "learning_rate": 9.06e-07, "loss": 0.0465, "num_tokens": 124344702.0, "reward": 0.193359375, "reward_std": 0.21742093563079834, "rewards/simpleverify_reward/mean": 0.193359375, "rewards/simpleverify_reward/std": 0.39531853795051575, "step": 48, "tools/generated_tokens": 3069.89453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.76171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.65234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1996.0, "completions/mean_length": 1716.580078125, "completions/mean_terminated_length": 1094.7022705078125, "completions/min_length": 303.0, "completions/min_terminated_length": 303.0, "epoch": 0.0347764371894961, "frac_reward_zero_std": 0.46875, "grad_norm": 6.799499034881592, "learning_rate": 9.039999999999999e-07, "loss": 0.0277, "num_tokens": 126895441.0, "reward": 0.19140625, "reward_std": 0.2129901945590973, "rewards/simpleverify_reward/mean": 0.19140625, "rewards/simpleverify_reward/std": 0.3937928080558777, "step": 49, "tools/generated_tokens": 3112.583984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.78125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.55078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1924.0, "completions/mean_length": 1633.029296875, "completions/mean_terminated_length": 1124.2391357421875, "completions/min_length": 356.0, "completions/min_terminated_length": 356.0, "epoch": 0.035486160397444996, "frac_reward_zero_std": 0.40625, "grad_norm": 22.707195281982422, "learning_rate": 9.02e-07, "loss": 0.0508, "num_tokens": 129399264.0, "reward": 0.275390625, "reward_std": 0.23492860794067383, "rewards/simpleverify_reward/mean": 0.275390625, "rewards/simpleverify_reward/std": 0.44714778661727905, "step": 50, "tools/generated_tokens": 2999.95703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.794921875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.5078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1975.0, "completions/mean_length": 1579.55859375, "completions/mean_terminated_length": 1097.0675048828125, "completions/min_length": 281.0, "completions/min_terminated_length": 281.0, "epoch": 0.0361958836053939, "frac_reward_zero_std": 0.34375, "grad_norm": 18.979333877563477, "learning_rate": 9e-07, "loss": 0.0366, "num_tokens": 131777685.0, "reward": 0.283203125, "reward_std": 0.2675431966781616, "rewards/simpleverify_reward/mean": 0.283203125, "rewards/simpleverify_reward/std": 0.4509948492050171, "step": 51, "tools/generated_tokens": 2842.3515625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.73046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.490234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1932.0, "completions/mean_length": 1590.4765625, "completions/mean_terminated_length": 1150.4827880859375, "completions/min_length": 288.0, "completions/min_terminated_length": 288.0, "epoch": 0.036905606813342796, "frac_reward_zero_std": 0.375, "grad_norm": 18.082664489746094, "learning_rate": 8.98e-07, "loss": 0.0562, "num_tokens": 134184578.0, "reward": 0.35546875, "reward_std": 0.23496931791305542, "rewards/simpleverify_reward/mean": 0.35546875, "rewards/simpleverify_reward/std": 0.47912323474884033, "step": 52, "tools/generated_tokens": 2907.470703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7421875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.57421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1918.0, "completions/mean_length": 1624.609375, "completions/mean_terminated_length": 1053.6146240234375, "completions/min_length": 235.0, "completions/min_terminated_length": 235.0, "epoch": 0.03761533002129169, "frac_reward_zero_std": 0.40625, "grad_norm": 6.872910976409912, "learning_rate": 8.96e-07, "loss": 0.0395, "num_tokens": 136559071.0, "reward": 0.2578125, "reward_std": 0.25343742966651917, "rewards/simpleverify_reward/mean": 0.2578125, "rewards/simpleverify_reward/std": 0.43785804510116577, "step": 53, "tools/generated_tokens": 2868.931640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.69921875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.556640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1928.0, "completions/mean_length": 1637.54296875, "completions/mean_terminated_length": 1124.696044921875, "completions/min_length": 363.0, "completions/min_terminated_length": 363.0, "epoch": 0.0383250532292406, "frac_reward_zero_std": 0.53125, "grad_norm": 28.91997718811035, "learning_rate": 8.939999999999999e-07, "loss": 0.0451, "num_tokens": 139057226.0, "reward": 0.3203125, "reward_std": 0.21085387468338013, "rewards/simpleverify_reward/mean": 0.3203125, "rewards/simpleverify_reward/std": 0.4670529365539551, "step": 54, "tools/generated_tokens": 3018.208984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.791015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.6640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1835.0, "completions/mean_length": 1765.376953125, "completions/mean_terminated_length": 1206.7034912109375, "completions/min_length": 577.0, "completions/min_terminated_length": 577.0, "epoch": 0.03903477643718949, "frac_reward_zero_std": 0.5625, "grad_norm": 9.452083587646484, "learning_rate": 8.92e-07, "loss": 0.0417, "num_tokens": 141670889.0, "reward": 0.259765625, "reward_std": 0.17298728227615356, "rewards/simpleverify_reward/mean": 0.259765625, "rewards/simpleverify_reward/std": 0.4389347732067108, "step": 55, "tools/generated_tokens": 3188.228515625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.79296875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.6015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1996.0, "completions/mean_length": 1665.740234375, "completions/mean_terminated_length": 1088.60302734375, "completions/min_length": 201.0, "completions/min_terminated_length": 201.0, "epoch": 0.0397444996451384, "frac_reward_zero_std": 0.4375, "grad_norm": 9.190205574035645, "learning_rate": 8.9e-07, "loss": 0.0446, "num_tokens": 144158962.0, "reward": 0.302734375, "reward_std": 0.22766923904418945, "rewards/simpleverify_reward/mean": 0.302734375, "rewards/simpleverify_reward/std": 0.45989060401916504, "step": 56, "tools/generated_tokens": 3050.205078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.76171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.732421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2007.0, "completions/mean_length": 1849.00390625, "completions/mean_terminated_length": 1304.3065185546875, "completions/min_length": 418.0, "completions/min_terminated_length": 418.0, "epoch": 0.04045422285308729, "frac_reward_zero_std": 0.53125, "grad_norm": 9.242337226867676, "learning_rate": 8.88e-07, "loss": 0.037, "num_tokens": 146924484.0, "reward": 0.22265625, "reward_std": 0.18859326839447021, "rewards/simpleverify_reward/mean": 0.22265625, "rewards/simpleverify_reward/std": 0.41643625497817993, "step": 57, "tools/generated_tokens": 3434.06640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.873046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.62890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1955.0, "completions/mean_length": 1720.654296875, "completions/mean_terminated_length": 1165.8895263671875, "completions/min_length": 307.0, "completions/min_terminated_length": 307.0, "epoch": 0.0411639460610362, "frac_reward_zero_std": 0.34375, "grad_norm": 8.084404945373535, "learning_rate": 8.86e-07, "loss": 0.0271, "num_tokens": 149498709.0, "reward": 0.244140625, "reward_std": 0.25465962290763855, "rewards/simpleverify_reward/mean": 0.244140625, "rewards/simpleverify_reward/std": 0.42999663949012756, "step": 58, "tools/generated_tokens": 3158.93359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.72265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1953.0, "completions/mean_length": 1807.2734375, "completions/mean_terminated_length": 1184.08447265625, "completions/min_length": 254.0, "completions/min_terminated_length": 254.0, "epoch": 0.04187366926898509, "frac_reward_zero_std": 0.625, "grad_norm": 3.5734550952911377, "learning_rate": 8.839999999999999e-07, "loss": 0.0112, "num_tokens": 152219043.0, "reward": 0.173828125, "reward_std": 0.13590145111083984, "rewards/simpleverify_reward/mean": 0.173828125, "rewards/simpleverify_reward/std": 0.3793322443962097, "step": 59, "tools/generated_tokens": 3337.78125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.861328125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.599609375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1924.0, "completions/mean_length": 1725.140625, "completions/mean_terminated_length": 1246.1658935546875, "completions/min_length": 449.0, "completions/min_terminated_length": 449.0, "epoch": 0.042583392476934, "frac_reward_zero_std": 0.25, "grad_norm": 22.264842987060547, "learning_rate": 8.82e-07, "loss": 0.0604, "num_tokens": 154844308.0, "reward": 0.28515625, "reward_std": 0.28179147839546204, "rewards/simpleverify_reward/mean": 0.28515625, "rewards/simpleverify_reward/std": 0.45193037390708923, "step": 60, "tools/generated_tokens": 3162.15234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.8203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.556640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2002.0, "completions/mean_length": 1717.20703125, "completions/mean_terminated_length": 1301.8941650390625, "completions/min_length": 375.0, "completions/min_terminated_length": 375.0, "epoch": 0.043293115684882894, "frac_reward_zero_std": 0.34375, "grad_norm": 15.20835018157959, "learning_rate": 8.799999999999999e-07, "loss": 0.0421, "num_tokens": 157391425.0, "reward": 0.275390625, "reward_std": 0.2693071961402893, "rewards/simpleverify_reward/mean": 0.275390625, "rewards/simpleverify_reward/std": 0.44714778661727905, "step": 61, "tools/generated_tokens": 3117.513671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.783203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.60546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1968.0, "completions/mean_length": 1744.361328125, "completions/mean_terminated_length": 1278.3812255859375, "completions/min_length": 305.0, "completions/min_terminated_length": 305.0, "epoch": 0.0440028388928318, "frac_reward_zero_std": 0.5, "grad_norm": 8.532882690429688, "learning_rate": 8.78e-07, "loss": 0.0434, "num_tokens": 160013624.0, "reward": 0.244140625, "reward_std": 0.21209990978240967, "rewards/simpleverify_reward/mean": 0.244140625, "rewards/simpleverify_reward/std": 0.42999663949012756, "step": 62, "tools/generated_tokens": 3206.244140625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.810546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.5078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1951.0, "completions/mean_length": 1573.0859375, "completions/mean_terminated_length": 1085.761962890625, "completions/min_length": 243.0, "completions/min_terminated_length": 243.0, "epoch": 0.044712562100780694, "frac_reward_zero_std": 0.28125, "grad_norm": 22.382343292236328, "learning_rate": 8.76e-07, "loss": 0.0558, "num_tokens": 162362298.0, "reward": 0.359375, "reward_std": 0.30179184675216675, "rewards/simpleverify_reward/mean": 0.359375, "rewards/simpleverify_reward/std": 0.48028653860092163, "step": 63, "tools/generated_tokens": 2813.314453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.728515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.76171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1935.0, "completions/mean_length": 1835.49609375, "completions/mean_terminated_length": 1165.91796875, "completions/min_length": 323.0, "completions/min_terminated_length": 323.0, "epoch": 0.0454222853087296, "frac_reward_zero_std": 0.5625, "grad_norm": 5.332128047943115, "learning_rate": 8.739999999999999e-07, "loss": 0.0352, "num_tokens": 165116312.0, "reward": 0.181640625, "reward_std": 0.16663315892219543, "rewards/simpleverify_reward/mean": 0.181640625, "rewards/simpleverify_reward/std": 0.38592514395713806, "step": 64, "tools/generated_tokens": 3400.265625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.873046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.5234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1948.0, "completions/mean_length": 1649.2265625, "completions/mean_terminated_length": 1211.2294921875, "completions/min_length": 236.0, "completions/min_terminated_length": 236.0, "epoch": 0.046132008516678494, "frac_reward_zero_std": 0.3125, "grad_norm": 10.27473258972168, "learning_rate": 8.72e-07, "loss": 0.0713, "num_tokens": 167543390.0, "reward": 0.3125, "reward_std": 0.2862105369567871, "rewards/simpleverify_reward/mean": 0.3125, "rewards/simpleverify_reward/std": 0.4639657139778137, "step": 65, "tools/generated_tokens": 2976.14453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.74609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.685546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1960.0, "completions/mean_length": 1755.50390625, "completions/mean_terminated_length": 1117.8260498046875, "completions/min_length": 382.0, "completions/min_terminated_length": 382.0, "epoch": 0.0468417317246274, "frac_reward_zero_std": 0.5, "grad_norm": 14.924501419067383, "learning_rate": 8.699999999999999e-07, "loss": 0.0372, "num_tokens": 170083724.0, "reward": 0.28515625, "reward_std": 0.17642264068126678, "rewards/simpleverify_reward/mean": 0.28515625, "rewards/simpleverify_reward/std": 0.45193037390708923, "step": 66, "tools/generated_tokens": 3101.615234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.771484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.59765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1978.0, "completions/mean_length": 1719.275390625, "completions/mean_terminated_length": 1230.9757080078125, "completions/min_length": 359.0, "completions/min_terminated_length": 359.0, "epoch": 0.047551454932576294, "frac_reward_zero_std": 0.34375, "grad_norm": 13.187244415283203, "learning_rate": 8.68e-07, "loss": 0.0603, "num_tokens": 172701987.0, "reward": 0.27734375, "reward_std": 0.26239949464797974, "rewards/simpleverify_reward/mean": 0.27734375, "rewards/simpleverify_reward/std": 0.4481254518032074, "step": 67, "tools/generated_tokens": 3196.9765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.818359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.505859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1975.0, "completions/mean_length": 1715.0, "completions/mean_terminated_length": 1374.102783203125, "completions/min_length": 395.0, "completions/min_terminated_length": 395.0, "epoch": 0.0482611781405252, "frac_reward_zero_std": 0.40625, "grad_norm": 9.947080612182617, "learning_rate": 8.659999999999999e-07, "loss": 0.046, "num_tokens": 175314953.0, "reward": 0.337890625, "reward_std": 0.25332921743392944, "rewards/simpleverify_reward/mean": 0.337890625, "rewards/simpleverify_reward/std": 0.4734536409378052, "step": 68, "tools/generated_tokens": 3209.82421875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.830078125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.65625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1979.0, "completions/mean_length": 1791.228515625, "completions/mean_terminated_length": 1301.0284423828125, "completions/min_length": 444.0, "completions/min_terminated_length": 444.0, "epoch": 0.048970901348474094, "frac_reward_zero_std": 0.4375, "grad_norm": 6.0258989334106445, "learning_rate": 8.639999999999999e-07, "loss": 0.0415, "num_tokens": 178040084.0, "reward": 0.2109375, "reward_std": 0.23673264682292938, "rewards/simpleverify_reward/mean": 0.2109375, "rewards/simpleverify_reward/std": 0.4083731174468994, "step": 69, "tools/generated_tokens": 3360.18359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.88671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.5546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1945.0, "completions/mean_length": 1686.498046875, "completions/mean_terminated_length": 1236.2061767578125, "completions/min_length": 317.0, "completions/min_terminated_length": 317.0, "epoch": 0.049680624556423, "frac_reward_zero_std": 0.40625, "grad_norm": 5.906918048858643, "learning_rate": 8.62e-07, "loss": 0.057, "num_tokens": 180509021.0, "reward": 0.255859375, "reward_std": 0.21704861521720886, "rewards/simpleverify_reward/mean": 0.255859375, "rewards/simpleverify_reward/std": 0.43676990270614624, "step": 70, "tools/generated_tokens": 2996.49609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.576171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1978.0, "completions/mean_length": 1681.802734375, "completions/mean_terminated_length": 1184.10595703125, "completions/min_length": 261.0, "completions/min_terminated_length": 261.0, "epoch": 0.050390347764371894, "frac_reward_zero_std": 0.375, "grad_norm": 10.25905704498291, "learning_rate": 8.599999999999999e-07, "loss": 0.0402, "num_tokens": 183007500.0, "reward": 0.255859375, "reward_std": 0.2586750388145447, "rewards/simpleverify_reward/mean": 0.255859375, "rewards/simpleverify_reward/std": 0.43676990270614624, "step": 71, "tools/generated_tokens": 3028.48828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.76171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1972.0, "completions/mean_length": 1746.3984375, "completions/mean_terminated_length": 1246.244873046875, "completions/min_length": 316.0, "completions/min_terminated_length": 316.0, "epoch": 0.0511000709723208, "frac_reward_zero_std": 0.40625, "grad_norm": 5.019900321960449, "learning_rate": 8.58e-07, "loss": 0.0557, "num_tokens": 185657170.0, "reward": 0.30859375, "reward_std": 0.24284851551055908, "rewards/simpleverify_reward/mean": 0.30859375, "rewards/simpleverify_reward/std": 0.4623647928237915, "step": 72, "tools/generated_tokens": 3279.6796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.83984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.572265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1960.0, "completions/mean_length": 1665.2578125, "completions/mean_terminated_length": 1153.1871337890625, "completions/min_length": 295.0, "completions/min_terminated_length": 295.0, "epoch": 0.051809794180269694, "frac_reward_zero_std": 0.4375, "grad_norm": 3.6638171672821045, "learning_rate": 8.559999999999999e-07, "loss": 0.0409, "num_tokens": 188021101.0, "reward": 0.298828125, "reward_std": 0.21144573390483856, "rewards/simpleverify_reward/mean": 0.298828125, "rewards/simpleverify_reward/std": 0.45819199085235596, "step": 73, "tools/generated_tokens": 2882.314453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.673828125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.59375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1963.0, "completions/mean_length": 1707.26953125, "completions/mean_terminated_length": 1213.75, "completions/min_length": 395.0, "completions/min_terminated_length": 395.0, "epoch": 0.0525195173882186, "frac_reward_zero_std": 0.46875, "grad_norm": 3.2305357456207275, "learning_rate": 8.539999999999999e-07, "loss": 0.0387, "num_tokens": 190566393.0, "reward": 0.185546875, "reward_std": 0.23356732726097107, "rewards/simpleverify_reward/mean": 0.185546875, "rewards/simpleverify_reward/std": 0.38912075757980347, "step": 74, "tools/generated_tokens": 3132.56640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.78125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.490234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1946.0, "completions/mean_length": 1598.220703125, "completions/mean_terminated_length": 1166.4405517578125, "completions/min_length": 272.0, "completions/min_terminated_length": 272.0, "epoch": 0.053229240596167494, "frac_reward_zero_std": 0.28125, "grad_norm": 8.233427047729492, "learning_rate": 8.52e-07, "loss": 0.0368, "num_tokens": 192970137.0, "reward": 0.392578125, "reward_std": 0.28937748074531555, "rewards/simpleverify_reward/mean": 0.392578125, "rewards/simpleverify_reward/std": 0.4888018071651459, "step": 75, "tools/generated_tokens": 2911.8984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.759765625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.62109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1947.0, "completions/mean_length": 1704.2109375, "completions/mean_terminated_length": 1140.680419921875, "completions/min_length": 346.0, "completions/min_terminated_length": 346.0, "epoch": 0.0539389638041164, "frac_reward_zero_std": 0.65625, "grad_norm": 8.22058391571045, "learning_rate": 8.499999999999999e-07, "loss": 0.0242, "num_tokens": 195477206.0, "reward": 0.255859375, "reward_std": 0.1335323303937912, "rewards/simpleverify_reward/mean": 0.255859375, "rewards/simpleverify_reward/std": 0.43676990270614624, "step": 76, "tools/generated_tokens": 3056.87109375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.751953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.544921875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1942.0, "completions/mean_length": 1602.326171875, "completions/mean_terminated_length": 1068.665283203125, "completions/min_length": 199.0, "completions/min_terminated_length": 199.0, "epoch": 0.054648687012065295, "frac_reward_zero_std": 0.5, "grad_norm": 3.9568536281585693, "learning_rate": 8.48e-07, "loss": 0.0217, "num_tokens": 197849791.0, "reward": 0.2421875, "reward_std": 0.20984011888504028, "rewards/simpleverify_reward/mean": 0.2421875, "rewards/simpleverify_reward/std": 0.42882615327835083, "step": 77, "tools/generated_tokens": 2839.375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.71484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.529296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2000.0, "completions/mean_length": 1664.63671875, "completions/mean_terminated_length": 1236.10791015625, "completions/min_length": 377.0, "completions/min_terminated_length": 377.0, "epoch": 0.05535841022001419, "frac_reward_zero_std": 0.34375, "grad_norm": 3.7316529750823975, "learning_rate": 8.459999999999999e-07, "loss": 0.039, "num_tokens": 200335246.0, "reward": 0.38671875, "reward_std": 0.26351770758628845, "rewards/simpleverify_reward/mean": 0.38671875, "rewards/simpleverify_reward/std": 0.48747459053993225, "step": 78, "tools/generated_tokens": 3043.78515625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.505859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1987.0, "completions/mean_length": 1577.09375, "completions/mean_terminated_length": 1095.019775390625, "completions/min_length": 282.0, "completions/min_terminated_length": 282.0, "epoch": 0.056068133427963095, "frac_reward_zero_std": 0.34375, "grad_norm": 17.494464874267578, "learning_rate": 8.439999999999999e-07, "loss": 0.0265, "num_tokens": 202684836.0, "reward": 0.341796875, "reward_std": 0.25113826990127563, "rewards/simpleverify_reward/mean": 0.341796875, "rewards/simpleverify_reward/std": 0.4747757613658905, "step": 79, "tools/generated_tokens": 2793.693359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.638671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1999.0, "completions/mean_length": 1734.890625, "completions/mean_terminated_length": 1184.54052734375, "completions/min_length": 389.0, "completions/min_terminated_length": 389.0, "epoch": 0.05677785663591199, "frac_reward_zero_std": 0.40625, "grad_norm": 4.200736999511719, "learning_rate": 8.419999999999999e-07, "loss": 0.0369, "num_tokens": 205300898.0, "reward": 0.263671875, "reward_std": 0.22413700819015503, "rewards/simpleverify_reward/mean": 0.263671875, "rewards/simpleverify_reward/std": 0.4410543739795685, "step": 80, "tools/generated_tokens": 3202.73828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.818359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.5703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1965.0, "completions/mean_length": 1670.69140625, "completions/mean_terminated_length": 1170.8863525390625, "completions/min_length": 327.0, "completions/min_terminated_length": 327.0, "epoch": 0.057487579843860895, "frac_reward_zero_std": 0.34375, "grad_norm": 3.668684959411621, "learning_rate": 8.399999999999999e-07, "loss": 0.0376, "num_tokens": 207803671.0, "reward": 0.302734375, "reward_std": 0.24782003462314606, "rewards/simpleverify_reward/mean": 0.302734375, "rewards/simpleverify_reward/std": 0.45989060401916504, "step": 81, "tools/generated_tokens": 3057.98046875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.775390625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1964.0, "completions/mean_length": 1788.927734375, "completions/mean_terminated_length": 1258.4464111328125, "completions/min_length": 298.0, "completions/min_terminated_length": 298.0, "epoch": 0.05819730305180979, "frac_reward_zero_std": 0.46875, "grad_norm": 10.686431884765625, "learning_rate": 8.38e-07, "loss": 0.0464, "num_tokens": 210533270.0, "reward": 0.205078125, "reward_std": 0.21273121237754822, "rewards/simpleverify_reward/mean": 0.205078125, "rewards/simpleverify_reward/std": 0.4041535556316376, "step": 82, "tools/generated_tokens": 3339.1328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.560546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1967.0, "completions/mean_length": 1665.58984375, "completions/mean_terminated_length": 1177.804443359375, "completions/min_length": 147.0, "completions/min_terminated_length": 147.0, "epoch": 0.058907026259758695, "frac_reward_zero_std": 0.4375, "grad_norm": 4.3856401443481445, "learning_rate": 8.359999999999999e-07, "loss": 0.0555, "num_tokens": 212998267.0, "reward": 0.337890625, "reward_std": 0.2313145101070404, "rewards/simpleverify_reward/mean": 0.337890625, "rewards/simpleverify_reward/std": 0.4734536409378052, "step": 83, "tools/generated_tokens": 3019.388671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.744140625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.431640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1914.0, "completions/mean_length": 1473.767578125, "completions/mean_terminated_length": 1038.7388916015625, "completions/min_length": 235.0, "completions/min_terminated_length": 235.0, "epoch": 0.05961674946770759, "frac_reward_zero_std": 0.28125, "grad_norm": 6.653895378112793, "learning_rate": 8.34e-07, "loss": 0.1007, "num_tokens": 215188589.0, "reward": 0.3984375, "reward_std": 0.30831027030944824, "rewards/simpleverify_reward/mean": 0.3984375, "rewards/simpleverify_reward/std": 0.4900552034378052, "step": 84, "tools/generated_tokens": 2597.296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.63671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.65234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1980.0, "completions/mean_length": 1736.013671875, "completions/mean_terminated_length": 1150.994384765625, "completions/min_length": 246.0, "completions/min_terminated_length": 246.0, "epoch": 0.060326472675656495, "frac_reward_zero_std": 0.53125, "grad_norm": 1.9694050550460815, "learning_rate": 8.319999999999999e-07, "loss": 0.0228, "num_tokens": 217801939.0, "reward": 0.189453125, "reward_std": 0.18039482831954956, "rewards/simpleverify_reward/mean": 0.189453125, "rewards/simpleverify_reward/std": 0.3922513723373413, "step": 85, "tools/generated_tokens": 3218.017578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.81640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.669921875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1939.0, "completions/mean_length": 1680.2734375, "completions/mean_terminated_length": 945.6567993164062, "completions/min_length": 258.0, "completions/min_terminated_length": 258.0, "epoch": 0.06103619588360539, "frac_reward_zero_std": 0.46875, "grad_norm": 5.795001983642578, "learning_rate": 8.299999999999999e-07, "loss": 0.0268, "num_tokens": 220340496.0, "reward": 0.208984375, "reward_std": 0.19692271947860718, "rewards/simpleverify_reward/mean": 0.208984375, "rewards/simpleverify_reward/std": 0.40698084235191345, "step": 86, "tools/generated_tokens": 3103.787109375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.8046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.595703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1951.0, "completions/mean_length": 1686.263671875, "completions/mean_terminated_length": 1153.2705078125, "completions/min_length": 331.0, "completions/min_terminated_length": 331.0, "epoch": 0.061745919091554295, "frac_reward_zero_std": 0.28125, "grad_norm": 7.0717902183532715, "learning_rate": 8.28e-07, "loss": 0.0561, "num_tokens": 222899225.0, "reward": 0.2421875, "reward_std": 0.24356047809123993, "rewards/simpleverify_reward/mean": 0.2421875, "rewards/simpleverify_reward/std": 0.42882615327835083, "step": 87, "tools/generated_tokens": 3078.25390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.80078125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.439453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1967.0, "completions/mean_length": 1502.05078125, "completions/mean_terminated_length": 1078.77001953125, "completions/min_length": 250.0, "completions/min_terminated_length": 250.0, "epoch": 0.06245564229950319, "frac_reward_zero_std": 0.34375, "grad_norm": 17.3853702545166, "learning_rate": 8.259999999999999e-07, "loss": 0.0588, "num_tokens": 225077902.0, "reward": 0.396484375, "reward_std": 0.25027522444725037, "rewards/simpleverify_reward/mean": 0.396484375, "rewards/simpleverify_reward/std": 0.4896455705165863, "step": 88, "tools/generated_tokens": 2518.818359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.62890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.58203125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1987.0, "completions/mean_length": 1677.4453125, "completions/mean_terminated_length": 1163.4813232421875, "completions/min_length": 399.0, "completions/min_terminated_length": 399.0, "epoch": 0.0631653655074521, "frac_reward_zero_std": 0.46875, "grad_norm": 3.0299954414367676, "learning_rate": 8.24e-07, "loss": 0.0368, "num_tokens": 227565916.0, "reward": 0.318359375, "reward_std": 0.20984143018722534, "rewards/simpleverify_reward/mean": 0.318359375, "rewards/simpleverify_reward/std": 0.46629536151885986, "step": 89, "tools/generated_tokens": 3011.517578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.755859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.513671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1998.0, "completions/mean_length": 1536.931640625, "completions/mean_terminated_length": 1000.1325073242188, "completions/min_length": 331.0, "completions/min_terminated_length": 331.0, "epoch": 0.063875088715401, "frac_reward_zero_std": 0.40625, "grad_norm": 3.3232064247131348, "learning_rate": 8.219999999999999e-07, "loss": 0.0448, "num_tokens": 229830873.0, "reward": 0.365234375, "reward_std": 0.21005308628082275, "rewards/simpleverify_reward/mean": 0.365234375, "rewards/simpleverify_reward/std": 0.4819667339324951, "step": 90, "tools/generated_tokens": 2706.556640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.654296875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.51171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1971.0, "completions/mean_length": 1612.70703125, "completions/mean_terminated_length": 1157.1361083984375, "completions/min_length": 394.0, "completions/min_terminated_length": 394.0, "epoch": 0.06458481192334989, "frac_reward_zero_std": 0.3125, "grad_norm": 11.046449661254883, "learning_rate": 8.199999999999999e-07, "loss": 0.0562, "num_tokens": 232246955.0, "reward": 0.330078125, "reward_std": 0.27328845858573914, "rewards/simpleverify_reward/mean": 0.330078125, "rewards/simpleverify_reward/std": 0.47070086002349854, "step": 91, "tools/generated_tokens": 2866.21484375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.728515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.57421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1991.0, "completions/mean_length": 1719.3203125, "completions/mean_terminated_length": 1276.054931640625, "completions/min_length": 290.0, "completions/min_terminated_length": 290.0, "epoch": 0.06529453513129879, "frac_reward_zero_std": 0.3125, "grad_norm": 4.098525524139404, "learning_rate": 8.179999999999999e-07, "loss": 0.0656, "num_tokens": 234831740.0, "reward": 0.296875, "reward_std": 0.26665839552879333, "rewards/simpleverify_reward/mean": 0.296875, "rewards/simpleverify_reward/std": 0.45732781291007996, "step": 92, "tools/generated_tokens": 3155.52734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.810546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.40234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1964.0, "completions/mean_length": 1478.431640625, "completions/mean_terminated_length": 1096.61767578125, "completions/min_length": 179.0, "completions/min_terminated_length": 179.0, "epoch": 0.0660042583392477, "frac_reward_zero_std": 0.375, "grad_norm": 5.00880765914917, "learning_rate": 8.159999999999999e-07, "loss": 0.0568, "num_tokens": 237029456.0, "reward": 0.45703125, "reward_std": 0.2547861933708191, "rewards/simpleverify_reward/mean": 0.45703125, "rewards/simpleverify_reward/std": 0.49863746762275696, "step": 93, "tools/generated_tokens": 2635.611328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.64453125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.5390625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1963.0, "completions/mean_length": 1620.912109375, "completions/mean_terminated_length": 1121.4364013671875, "completions/min_length": 404.0, "completions/min_terminated_length": 404.0, "epoch": 0.0667139815471966, "frac_reward_zero_std": 0.34375, "grad_norm": 11.68735408782959, "learning_rate": 8.14e-07, "loss": 0.0567, "num_tokens": 239469666.0, "reward": 0.330078125, "reward_std": 0.25644606351852417, "rewards/simpleverify_reward/mean": 0.330078125, "rewards/simpleverify_reward/std": 0.47070086002349854, "step": 94, "tools/generated_tokens": 2971.802734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.765625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.00390625 }, { "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": 0.5546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1904.0, "completions/mean_length": 1611.423828125, "completions/mean_terminated_length": 1071.5614013671875, "completions/min_length": 227.0, "completions/min_terminated_length": 227.0, "epoch": 0.06742370475514549, "frac_reward_zero_std": 0.34375, "grad_norm": 3.146031618118286, "learning_rate": 8.12e-07, "loss": 0.0495, "num_tokens": 241876909.0, "reward": 0.298828125, "reward_std": 0.2574097812175751, "rewards/simpleverify_reward/mean": 0.298828125, "rewards/simpleverify_reward/std": 0.45819199085235596, "step": 95, "tools/generated_tokens": 2928.177734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.72265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.5234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1999.0, "completions/mean_length": 1656.95703125, "completions/mean_terminated_length": 1228.3072509765625, "completions/min_length": 443.0, "completions/min_terminated_length": 443.0, "epoch": 0.06813342796309439, "frac_reward_zero_std": 0.4375, "grad_norm": 8.736165046691895, "learning_rate": 8.1e-07, "loss": 0.0413, "num_tokens": 244378694.0, "reward": 0.333984375, "reward_std": 0.2218313366174698, "rewards/simpleverify_reward/mean": 0.333984375, "rewards/simpleverify_reward/std": 0.47209542989730835, "step": 96, "tools/generated_tokens": 3015.033203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.76171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.5078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1989.0, "completions/mean_length": 1550.619140625, "completions/mean_terminated_length": 1037.448486328125, "completions/min_length": 263.0, "completions/min_terminated_length": 263.0, "epoch": 0.0688431511710433, "frac_reward_zero_std": 0.4375, "grad_norm": 27.84614372253418, "learning_rate": 8.08e-07, "loss": 0.027, "num_tokens": 246695353.0, "reward": 0.294921875, "reward_std": 0.2198038101196289, "rewards/simpleverify_reward/mean": 0.294921875, "rewards/simpleverify_reward/std": 0.4564536213874817, "step": 97, "tools/generated_tokens": 2769.880859375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.697265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.521484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1981.0, "completions/mean_length": 1524.875, "completions/mean_terminated_length": 958.5877075195312, "completions/min_length": 294.0, "completions/min_terminated_length": 294.0, "epoch": 0.0695528743789922, "frac_reward_zero_std": 0.4375, "grad_norm": 4.464077949523926, "learning_rate": 8.06e-07, "loss": 0.0434, "num_tokens": 248968114.0, "reward": 0.26953125, "reward_std": 0.21846505999565125, "rewards/simpleverify_reward/mean": 0.26953125, "rewards/simpleverify_reward/std": 0.44415023922920227, "step": 98, "tools/generated_tokens": 2720.962890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.681640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.529296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1988.0, "completions/mean_length": 1620.564453125, "completions/mean_terminated_length": 1139.9212646484375, "completions/min_length": 181.0, "completions/min_terminated_length": 181.0, "epoch": 0.07026259758694109, "frac_reward_zero_std": 0.34375, "grad_norm": 7.144275665283203, "learning_rate": 8.04e-07, "loss": 0.037, "num_tokens": 251447379.0, "reward": 0.3203125, "reward_std": 0.2558642029762268, "rewards/simpleverify_reward/mean": 0.3203125, "rewards/simpleverify_reward/std": 0.4670529365539551, "step": 99, "tools/generated_tokens": 2979.216796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.802734375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.408203125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1958.0, "completions/mean_length": 1443.998046875, "completions/mean_terminated_length": 1027.376220703125, "completions/min_length": 262.0, "completions/min_terminated_length": 262.0, "epoch": 0.07097232079488999, "frac_reward_zero_std": 0.4375, "grad_norm": 2.353195905685425, "learning_rate": 8.02e-07, "loss": 0.0016, "num_tokens": 253543647.0, "reward": 0.291015625, "reward_std": 0.21078607439994812, "rewards/simpleverify_reward/mean": 0.291015625, "rewards/simpleverify_reward/std": 0.45467492938041687, "step": 100, "tools/generated_tokens": 2502.7890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.603515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1941.0, "completions/mean_length": 1695.72265625, "completions/mean_terminated_length": 1159.49755859375, "completions/min_length": 181.0, "completions/min_terminated_length": 181.0, "epoch": 0.0716820440028389, "frac_reward_zero_std": 0.375, "grad_norm": 7.572436332702637, "learning_rate": 8e-07, "loss": 0.0306, "num_tokens": 256025254.0, "reward": 0.28515625, "reward_std": 0.22731757164001465, "rewards/simpleverify_reward/mean": 0.28515625, "rewards/simpleverify_reward/std": 0.45193037390708923, "step": 101, "tools/generated_tokens": 3017.451171875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.732421875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.462890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1970.0, "completions/mean_length": 1552.8046875, "completions/mean_terminated_length": 1126.036376953125, "completions/min_length": 341.0, "completions/min_terminated_length": 341.0, "epoch": 0.0723917672107878, "frac_reward_zero_std": 0.34375, "grad_norm": 2.97236967086792, "learning_rate": 7.98e-07, "loss": 0.057, "num_tokens": 258294355.0, "reward": 0.431640625, "reward_std": 0.2626723647117615, "rewards/simpleverify_reward/mean": 0.431640625, "rewards/simpleverify_reward/std": 0.4957893490791321, "step": 102, "tools/generated_tokens": 2756.841796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.658203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.40625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1926.0, "completions/mean_length": 1430.443359375, "completions/mean_terminated_length": 1007.9046020507812, "completions/min_length": 225.0, "completions/min_terminated_length": 225.0, "epoch": 0.07310149041873669, "frac_reward_zero_std": 0.34375, "grad_norm": 15.120314598083496, "learning_rate": 7.96e-07, "loss": 0.0691, "num_tokens": 260395381.0, "reward": 0.4453125, "reward_std": 0.28424036502838135, "rewards/simpleverify_reward/mean": 0.4453125, "rewards/simpleverify_reward/std": 0.49748632311820984, "step": 103, "tools/generated_tokens": 2482.310546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.59765625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.5625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1995.0, "completions/mean_length": 1644.595703125, "completions/mean_terminated_length": 1125.93310546875, "completions/min_length": 142.0, "completions/min_terminated_length": 142.0, "epoch": 0.07381121362668559, "frac_reward_zero_std": 0.4375, "grad_norm": 10.363288879394531, "learning_rate": 7.94e-07, "loss": 0.0512, "num_tokens": 262803708.0, "reward": 0.3125, "reward_std": 0.23525568842887878, "rewards/simpleverify_reward/mean": 0.3125, "rewards/simpleverify_reward/std": 0.4639657139778137, "step": 104, "tools/generated_tokens": 2944.2890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.71484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.46875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1981.0, "completions/mean_length": 1548.431640625, "completions/mean_terminated_length": 1107.635986328125, "completions/min_length": 222.0, "completions/min_terminated_length": 222.0, "epoch": 0.0745209368346345, "frac_reward_zero_std": 0.375, "grad_norm": 5.745745658874512, "learning_rate": 7.92e-07, "loss": 0.0682, "num_tokens": 265121654.0, "reward": 0.375, "reward_std": 0.2560069262981415, "rewards/simpleverify_reward/mean": 0.375, "rewards/simpleverify_reward/std": 0.4845963716506958, "step": 105, "tools/generated_tokens": 2784.763671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.68359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.62109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1991.0, "completions/mean_length": 1695.14453125, "completions/mean_terminated_length": 1116.7525634765625, "completions/min_length": 323.0, "completions/min_terminated_length": 323.0, "epoch": 0.07523066004258339, "frac_reward_zero_std": 0.4375, "grad_norm": 4.311976432800293, "learning_rate": 7.9e-07, "loss": 0.0445, "num_tokens": 267633969.0, "reward": 0.279296875, "reward_std": 0.21847295761108398, "rewards/simpleverify_reward/mean": 0.279296875, "rewards/simpleverify_reward/std": 0.44909247756004333, "step": 106, "tools/generated_tokens": 3078.62890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.76953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.462890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1997.0, "completions/mean_length": 1598.8828125, "completions/mean_terminated_length": 1211.825439453125, "completions/min_length": 497.0, "completions/min_terminated_length": 497.0, "epoch": 0.07594038325053229, "frac_reward_zero_std": 0.28125, "grad_norm": 7.951910018920898, "learning_rate": 7.88e-07, "loss": 0.0433, "num_tokens": 269901011.0, "reward": 0.40625, "reward_std": 0.26898670196533203, "rewards/simpleverify_reward/mean": 0.40625, "rewards/simpleverify_reward/std": 0.49161264300346375, "step": 107, "tools/generated_tokens": 2743.078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.630859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.537109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1959.0, "completions/mean_length": 1656.951171875, "completions/mean_terminated_length": 1207.0421142578125, "completions/min_length": 328.0, "completions/min_terminated_length": 328.0, "epoch": 0.0766501064584812, "frac_reward_zero_std": 0.5, "grad_norm": 8.335955619812012, "learning_rate": 7.86e-07, "loss": 0.0475, "num_tokens": 272364031.0, "reward": 0.28125, "reward_std": 0.20871783792972565, "rewards/simpleverify_reward/mean": 0.28125, "rewards/simpleverify_reward/std": 0.45004892349243164, "step": 108, "tools/generated_tokens": 3023.3984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.638671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1958.0, "completions/mean_length": 1739.767578125, "completions/mean_terminated_length": 1194.9459228515625, "completions/min_length": 332.0, "completions/min_terminated_length": 332.0, "epoch": 0.0773598296664301, "frac_reward_zero_std": 0.46875, "grad_norm": 4.050718307495117, "learning_rate": 7.84e-07, "loss": 0.0341, "num_tokens": 274937120.0, "reward": 0.271484375, "reward_std": 0.2020200788974762, "rewards/simpleverify_reward/mean": 0.271484375, "rewards/simpleverify_reward/std": 0.44516023993492126, "step": 109, "tools/generated_tokens": 3173.5390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.587890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1977.0, "completions/mean_length": 1712.806640625, "completions/mean_terminated_length": 1234.639892578125, "completions/min_length": 330.0, "completions/min_terminated_length": 330.0, "epoch": 0.07806955287437899, "frac_reward_zero_std": 0.375, "grad_norm": 5.5220513343811035, "learning_rate": 7.82e-07, "loss": 0.0336, "num_tokens": 277520470.0, "reward": 0.3046875, "reward_std": 0.2346501350402832, "rewards/simpleverify_reward/mean": 0.3046875, "rewards/simpleverify_reward/std": 0.4607250988483429, "step": 110, "tools/generated_tokens": 3120.26953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.8125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.48828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1957.0, "completions/mean_length": 1631.087890625, "completions/mean_terminated_length": 1236.93505859375, "completions/min_length": 350.0, "completions/min_terminated_length": 350.0, "epoch": 0.07877927608232789, "frac_reward_zero_std": 0.3125, "grad_norm": 2.563208818435669, "learning_rate": 7.799999999999999e-07, "loss": 0.0567, "num_tokens": 279891480.0, "reward": 0.365234375, "reward_std": 0.2847670018672943, "rewards/simpleverify_reward/mean": 0.365234375, "rewards/simpleverify_reward/std": 0.4819667339324951, "step": 111, "tools/generated_tokens": 2913.0234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.701171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.47265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1999.0, "completions/mean_length": 1579.484375, "completions/mean_terminated_length": 1164.566650390625, "completions/min_length": 273.0, "completions/min_terminated_length": 273.0, "epoch": 0.0794889992902768, "frac_reward_zero_std": 0.3125, "grad_norm": 5.98987340927124, "learning_rate": 7.78e-07, "loss": 0.0242, "num_tokens": 282203759.0, "reward": 0.365234375, "reward_std": 0.2900904417037964, "rewards/simpleverify_reward/mean": 0.365234375, "rewards/simpleverify_reward/std": 0.4819667339324951, "step": 112, "tools/generated_tokens": 2817.685546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.46484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1982.0, "completions/mean_length": 1538.90234375, "completions/mean_terminated_length": 1100.149658203125, "completions/min_length": 158.0, "completions/min_terminated_length": 158.0, "epoch": 0.0801987224982257, "frac_reward_zero_std": 0.40625, "grad_norm": 8.211172103881836, "learning_rate": 7.76e-07, "loss": 0.0729, "num_tokens": 284479712.0, "reward": 0.416015625, "reward_std": 0.22766360640525818, "rewards/simpleverify_reward/mean": 0.416015625, "rewards/simpleverify_reward/std": 0.493378221988678, "step": 113, "tools/generated_tokens": 2717.091796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.544921875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1994.0, "completions/mean_length": 1674.3203125, "completions/mean_terminated_length": 1231.519287109375, "completions/min_length": 346.0, "completions/min_terminated_length": 346.0, "epoch": 0.08090844570617459, "frac_reward_zero_std": 0.34375, "grad_norm": 4.83544397354126, "learning_rate": 7.74e-07, "loss": 0.0545, "num_tokens": 287028146.0, "reward": 0.33203125, "reward_std": 0.28244972229003906, "rewards/simpleverify_reward/mean": 0.33203125, "rewards/simpleverify_reward/std": 0.47140273451805115, "step": 114, "tools/generated_tokens": 3118.609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.818359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.498046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1904.0, "completions/mean_length": 1621.16796875, "completions/mean_terminated_length": 1197.87548828125, "completions/min_length": 473.0, "completions/min_terminated_length": 473.0, "epoch": 0.08161816891412349, "frac_reward_zero_std": 0.40625, "grad_norm": 1.8935874700546265, "learning_rate": 7.72e-07, "loss": 0.0521, "num_tokens": 289340012.0, "reward": 0.37890625, "reward_std": 0.2183789610862732, "rewards/simpleverify_reward/mean": 0.37890625, "rewards/simpleverify_reward/std": 0.4855891764163971, "step": 115, "tools/generated_tokens": 2807.35546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.369140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2003.0, "completions/mean_length": 1438.310546875, "completions/mean_terminated_length": 1082.077392578125, "completions/min_length": 166.0, "completions/min_terminated_length": 166.0, "epoch": 0.0823278921220724, "frac_reward_zero_std": 0.28125, "grad_norm": 11.063997268676758, "learning_rate": 7.699999999999999e-07, "loss": 0.0357, "num_tokens": 291481581.0, "reward": 0.46484375, "reward_std": 0.31301459670066833, "rewards/simpleverify_reward/mean": 0.46484375, "rewards/simpleverify_reward/std": 0.49925029277801514, "step": 116, "tools/generated_tokens": 2536.59765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.62109375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.478515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1955.0, "completions/mean_length": 1614.234375, "completions/mean_terminated_length": 1217.5880126953125, "completions/min_length": 320.0, "completions/min_terminated_length": 320.0, "epoch": 0.0830376153300213, "frac_reward_zero_std": 0.25, "grad_norm": 3.8780367374420166, "learning_rate": 7.68e-07, "loss": 0.0516, "num_tokens": 293976072.0, "reward": 0.361328125, "reward_std": 0.30899620056152344, "rewards/simpleverify_reward/mean": 0.361328125, "rewards/simpleverify_reward/std": 0.48085519671440125, "step": 117, "tools/generated_tokens": 3053.330078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.79296875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.498046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1973.0, "completions/mean_length": 1609.80859375, "completions/mean_terminated_length": 1175.0272216796875, "completions/min_length": 308.0, "completions/min_terminated_length": 308.0, "epoch": 0.08374733853797019, "frac_reward_zero_std": 0.46875, "grad_norm": 2.8188023567199707, "learning_rate": 7.66e-07, "loss": 0.031, "num_tokens": 296352274.0, "reward": 0.3359375, "reward_std": 0.20950201153755188, "rewards/simpleverify_reward/mean": 0.3359375, "rewards/simpleverify_reward/std": 0.4727790653705597, "step": 118, "tools/generated_tokens": 2908.0234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.708984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.427734375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1951.0, "completions/mean_length": 1516.408203125, "completions/mean_terminated_length": 1126.6927490234375, "completions/min_length": 250.0, "completions/min_terminated_length": 250.0, "epoch": 0.08445706174591909, "frac_reward_zero_std": 0.25, "grad_norm": 5.207269191741943, "learning_rate": 7.64e-07, "loss": 0.0582, "num_tokens": 298574584.0, "reward": 0.3671875, "reward_std": 0.30044621229171753, "rewards/simpleverify_reward/mean": 0.3671875, "rewards/simpleverify_reward/std": 0.48250964283943176, "step": 119, "tools/generated_tokens": 2676.640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6328125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.5703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1994.0, "completions/mean_length": 1704.65234375, "completions/mean_terminated_length": 1251.5726318359375, "completions/min_length": 434.0, "completions/min_terminated_length": 434.0, "epoch": 0.085166784953868, "frac_reward_zero_std": 0.28125, "grad_norm": 5.407745838165283, "learning_rate": 7.62e-07, "loss": 0.0499, "num_tokens": 301168197.0, "reward": 0.328125, "reward_std": 0.2872231602668762, "rewards/simpleverify_reward/mean": 0.328125, "rewards/simpleverify_reward/std": 0.4699897766113281, "step": 120, "tools/generated_tokens": 3198.15234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.81640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.537109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1945.0, "completions/mean_length": 1615.9375, "completions/mean_terminated_length": 1114.59912109375, "completions/min_length": 158.0, "completions/min_terminated_length": 158.0, "epoch": 0.0858765081618169, "frac_reward_zero_std": 0.4375, "grad_norm": 4.495778560638428, "learning_rate": 7.599999999999999e-07, "loss": 0.0476, "num_tokens": 303624701.0, "reward": 0.306640625, "reward_std": 0.24884289503097534, "rewards/simpleverify_reward/mean": 0.306640625, "rewards/simpleverify_reward/std": 0.4615498185157776, "step": 121, "tools/generated_tokens": 2998.318359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.404296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1973.0, "completions/mean_length": 1503.783203125, "completions/mean_terminated_length": 1138.2393798828125, "completions/min_length": 178.0, "completions/min_terminated_length": 178.0, "epoch": 0.08658623136976579, "frac_reward_zero_std": 0.25, "grad_norm": 2.2481284141540527, "learning_rate": 7.58e-07, "loss": 0.0496, "num_tokens": 305814545.0, "reward": 0.419921875, "reward_std": 0.3178216814994812, "rewards/simpleverify_reward/mean": 0.419921875, "rewards/simpleverify_reward/std": 0.4940285086631775, "step": 122, "tools/generated_tokens": 2643.154296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.416015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1944.0, "completions/mean_length": 1576.75390625, "completions/mean_terminated_length": 1241.0501708984375, "completions/min_length": 390.0, "completions/min_terminated_length": 390.0, "epoch": 0.08729595457771469, "frac_reward_zero_std": 0.3125, "grad_norm": 5.253964900970459, "learning_rate": 7.559999999999999e-07, "loss": 0.0258, "num_tokens": 308117094.0, "reward": 0.306640625, "reward_std": 0.2990352511405945, "rewards/simpleverify_reward/mean": 0.306640625, "rewards/simpleverify_reward/std": 0.4615498185157776, "step": 123, "tools/generated_tokens": 2757.0, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.65234375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.353515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1979.0, "completions/mean_length": 1400.26171875, "completions/mean_terminated_length": 1050.3685302734375, "completions/min_length": 286.0, "completions/min_terminated_length": 286.0, "epoch": 0.0880056777856636, "frac_reward_zero_std": 0.21875, "grad_norm": 3.416583299636841, "learning_rate": 7.54e-07, "loss": 0.0884, "num_tokens": 310199910.0, "reward": 0.43359375, "reward_std": 0.3443273603916168, "rewards/simpleverify_reward/mean": 0.43359375, "rewards/simpleverify_reward/std": 0.4960552453994751, "step": 124, "tools/generated_tokens": 2476.638671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.599609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.505859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1992.0, "completions/mean_length": 1659.6875, "completions/mean_terminated_length": 1262.166015625, "completions/min_length": 263.0, "completions/min_terminated_length": 263.0, "epoch": 0.0887154009936125, "frac_reward_zero_std": 0.1875, "grad_norm": 4.123121738433838, "learning_rate": 7.52e-07, "loss": 0.08, "num_tokens": 312701963.0, "reward": 0.359375, "reward_std": 0.33996930718421936, "rewards/simpleverify_reward/mean": 0.359375, "rewards/simpleverify_reward/std": 0.48028653860092163, "step": 125, "tools/generated_tokens": 3075.64453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.775390625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.419921875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1961.0, "completions/mean_length": 1532.328125, "completions/mean_terminated_length": 1161.8787841796875, "completions/min_length": 331.0, "completions/min_terminated_length": 331.0, "epoch": 0.08942512420156139, "frac_reward_zero_std": 0.46875, "grad_norm": 2.2802138328552246, "learning_rate": 7.5e-07, "loss": 0.0341, "num_tokens": 314913907.0, "reward": 0.47265625, "reward_std": 0.2003580629825592, "rewards/simpleverify_reward/mean": 0.47265625, "rewards/simpleverify_reward/std": 0.49974003434181213, "step": 126, "tools/generated_tokens": 2676.392578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6328125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.443359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1955.0, "completions/mean_length": 1513.861328125, "completions/mean_terminated_length": 1088.424560546875, "completions/min_length": 457.0, "completions/min_terminated_length": 457.0, "epoch": 0.09013484740951029, "frac_reward_zero_std": 0.40625, "grad_norm": 1.5407387018203735, "learning_rate": 7.48e-07, "loss": 0.0465, "num_tokens": 317161825.0, "reward": 0.30078125, "reward_std": 0.26212018728256226, "rewards/simpleverify_reward/mean": 0.30078125, "rewards/simpleverify_reward/std": 0.45904624462127686, "step": 127, "tools/generated_tokens": 2683.6484375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.642578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.00390625 }, { "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": 0.455078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1986.0, "completions/mean_length": 1523.99609375, "completions/mean_terminated_length": 1086.3870849609375, "completions/min_length": 206.0, "completions/min_terminated_length": 206.0, "epoch": 0.0908445706174592, "frac_reward_zero_std": 0.4375, "grad_norm": 2.7554562091827393, "learning_rate": 7.459999999999999e-07, "loss": 0.0734, "num_tokens": 319388571.0, "reward": 0.458984375, "reward_std": 0.2304583191871643, "rewards/simpleverify_reward/mean": 0.458984375, "rewards/simpleverify_reward/std": 0.49880221486091614, "step": 128, "tools/generated_tokens": 2677.345703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.634765625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.556640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1911.0, "completions/mean_length": 1615.048828125, "completions/mean_terminated_length": 1071.4757080078125, "completions/min_length": 356.0, "completions/min_terminated_length": 356.0, "epoch": 0.09155429382540808, "frac_reward_zero_std": 0.53125, "grad_norm": 3.5665786266326904, "learning_rate": 7.44e-07, "loss": 0.0226, "num_tokens": 321778433.0, "reward": 0.267578125, "reward_std": 0.1967860758304596, "rewards/simpleverify_reward/mean": 0.267578125, "rewards/simpleverify_reward/std": 0.4431293308734894, "step": 129, "tools/generated_tokens": 2882.830078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.40234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1934.0, "completions/mean_length": 1509.076171875, "completions/mean_terminated_length": 1148.8333740234375, "completions/min_length": 305.0, "completions/min_terminated_length": 305.0, "epoch": 0.09226401703335699, "frac_reward_zero_std": 0.375, "grad_norm": 1.4158332347869873, "learning_rate": 7.42e-07, "loss": 0.0483, "num_tokens": 323948261.0, "reward": 0.373046875, "reward_std": 0.2554856240749359, "rewards/simpleverify_reward/mean": 0.373046875, "rewards/simpleverify_reward/std": 0.48408737778663635, "step": 130, "tools/generated_tokens": 2614.3046875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.60546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.322265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1994.0, "completions/mean_length": 1456.31640625, "completions/mean_terminated_length": 1176.5965576171875, "completions/min_length": 213.0, "completions/min_terminated_length": 213.0, "epoch": 0.09297374024130589, "frac_reward_zero_std": 0.21875, "grad_norm": 4.234163761138916, "learning_rate": 7.4e-07, "loss": 0.0519, "num_tokens": 326129351.0, "reward": 0.40234375, "reward_std": 0.3200967609882355, "rewards/simpleverify_reward/mean": 0.40234375, "rewards/simpleverify_reward/std": 0.4908501207828522, "step": 131, "tools/generated_tokens": 2628.111328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.63671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.451171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1988.0, "completions/mean_length": 1551.826171875, "completions/mean_terminated_length": 1144.0889892578125, "completions/min_length": 49.0, "completions/min_terminated_length": 49.0, "epoch": 0.0936834634492548, "frac_reward_zero_std": 0.34375, "grad_norm": 3.298482894897461, "learning_rate": 7.38e-07, "loss": 0.0624, "num_tokens": 328374800.0, "reward": 0.419921875, "reward_std": 0.2815532684326172, "rewards/simpleverify_reward/mean": 0.419921875, "rewards/simpleverify_reward/std": 0.4940285086631775, "step": 132, "tools/generated_tokens": 2701.6015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.53515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1968.0, "completions/mean_length": 1579.060546875, "completions/mean_terminated_length": 1041.81103515625, "completions/min_length": 133.0, "completions/min_terminated_length": 133.0, "epoch": 0.09439318665720368, "frac_reward_zero_std": 0.3125, "grad_norm": 4.151210784912109, "learning_rate": 7.359999999999999e-07, "loss": 0.0579, "num_tokens": 330791839.0, "reward": 0.29296875, "reward_std": 0.29341334104537964, "rewards/simpleverify_reward/mean": 0.29296875, "rewards/simpleverify_reward/std": 0.455569326877594, "step": 133, "tools/generated_tokens": 2919.32421875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.73046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.416015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2003.0, "completions/mean_length": 1542.576171875, "completions/mean_terminated_length": 1182.5250244140625, "completions/min_length": 200.0, "completions/min_terminated_length": 200.0, "epoch": 0.09510290986515259, "frac_reward_zero_std": 0.34375, "grad_norm": 3.1761157512664795, "learning_rate": 7.34e-07, "loss": 0.0451, "num_tokens": 333011474.0, "reward": 0.31640625, "reward_std": 0.2668527662754059, "rewards/simpleverify_reward/mean": 0.31640625, "rewards/simpleverify_reward/std": 0.46552830934524536, "step": 134, "tools/generated_tokens": 2665.375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.470703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1984.0, "completions/mean_length": 1568.66015625, "completions/mean_terminated_length": 1142.3837890625, "completions/min_length": 344.0, "completions/min_terminated_length": 344.0, "epoch": 0.09581263307310149, "frac_reward_zero_std": 0.3125, "grad_norm": 4.591899394989014, "learning_rate": 7.319999999999999e-07, "loss": 0.04, "num_tokens": 335355252.0, "reward": 0.3359375, "reward_std": 0.26223433017730713, "rewards/simpleverify_reward/mean": 0.3359375, "rewards/simpleverify_reward/std": 0.4727790653705597, "step": 135, "tools/generated_tokens": 2812.376953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.009765625 }, { "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": 0.681640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1934.0, "completions/mean_length": 1771.806640625, "completions/mean_terminated_length": 1180.4478759765625, "completions/min_length": 353.0, "completions/min_terminated_length": 353.0, "epoch": 0.0965223562810504, "frac_reward_zero_std": 0.4375, "grad_norm": 3.6996829509735107, "learning_rate": 7.3e-07, "loss": 0.0263, "num_tokens": 338032016.0, "reward": 0.18359375, "reward_std": 0.21973475813865662, "rewards/simpleverify_reward/mean": 0.18359375, "rewards/simpleverify_reward/std": 0.3875311613082886, "step": 136, "tools/generated_tokens": 3309.642578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.845703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.556640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1997.0, "completions/mean_length": 1650.009765625, "completions/mean_terminated_length": 1150.330322265625, "completions/min_length": 282.0, "completions/min_terminated_length": 282.0, "epoch": 0.09723207948899928, "frac_reward_zero_std": 0.40625, "grad_norm": 2.863053560256958, "learning_rate": 7.28e-07, "loss": 0.0477, "num_tokens": 340543201.0, "reward": 0.2734375, "reward_std": 0.2658785879611969, "rewards/simpleverify_reward/mean": 0.2734375, "rewards/simpleverify_reward/std": 0.4461594223976135, "step": 137, "tools/generated_tokens": 3054.8359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.76953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.591796875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1994.0, "completions/mean_length": 1753.294921875, "completions/mean_terminated_length": 1331.64111328125, "completions/min_length": 64.0, "completions/min_terminated_length": 64.0, "epoch": 0.09794180269694819, "frac_reward_zero_std": 0.375, "grad_norm": 4.595541477203369, "learning_rate": 7.259999999999999e-07, "loss": 0.0673, "num_tokens": 343129934.0, "reward": 0.302734375, "reward_std": 0.2633458375930786, "rewards/simpleverify_reward/mean": 0.302734375, "rewards/simpleverify_reward/std": 0.45989060401916504, "step": 138, "tools/generated_tokens": 3200.580078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.79296875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1976.0, "completions/mean_length": 1677.4453125, "completions/mean_terminated_length": 1230.22412109375, "completions/min_length": 459.0, "completions/min_terminated_length": 459.0, "epoch": 0.09865152590489709, "frac_reward_zero_std": 0.40625, "grad_norm": 4.288841247558594, "learning_rate": 7.24e-07, "loss": 0.041, "num_tokens": 345619204.0, "reward": 0.234375, "reward_std": 0.23609964549541473, "rewards/simpleverify_reward/mean": 0.234375, "rewards/simpleverify_reward/std": 0.42402184009552, "step": 139, "tools/generated_tokens": 3046.857421875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.751953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.35546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1999.0, "completions/mean_length": 1476.1953125, "completions/mean_terminated_length": 1160.8363037109375, "completions/min_length": 246.0, "completions/min_terminated_length": 246.0, "epoch": 0.099361249112846, "frac_reward_zero_std": 0.4375, "grad_norm": 3.0428848266601562, "learning_rate": 7.219999999999999e-07, "loss": 0.0269, "num_tokens": 347809869.0, "reward": 0.404296875, "reward_std": 0.2513276934623718, "rewards/simpleverify_reward/mean": 0.404296875, "rewards/simpleverify_reward/std": 0.4912354052066803, "step": 140, "tools/generated_tokens": 2624.5234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.623046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.6015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1952.0, "completions/mean_length": 1714.701171875, "completions/mean_terminated_length": 1214.6422119140625, "completions/min_length": 349.0, "completions/min_terminated_length": 349.0, "epoch": 0.10007097232079488, "frac_reward_zero_std": 0.28125, "grad_norm": 4.586359977722168, "learning_rate": 7.2e-07, "loss": 0.0634, "num_tokens": 350381406.0, "reward": 0.26953125, "reward_std": 0.2809322476387024, "rewards/simpleverify_reward/mean": 0.26953125, "rewards/simpleverify_reward/std": 0.44415023922920227, "step": 141, "tools/generated_tokens": 3164.609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.794921875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.419921875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1942.0, "completions/mean_length": 1482.53515625, "completions/mean_terminated_length": 1075.4140625, "completions/min_length": 235.0, "completions/min_terminated_length": 235.0, "epoch": 0.10078069552874379, "frac_reward_zero_std": 0.46875, "grad_norm": 2.6151556968688965, "learning_rate": 7.179999999999999e-07, "loss": 0.0185, "num_tokens": 352550499.0, "reward": 0.380859375, "reward_std": 0.22061602771282196, "rewards/simpleverify_reward/mean": 0.380859375, "rewards/simpleverify_reward/std": 0.48607301712036133, "step": 142, "tools/generated_tokens": 2599.20703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.623046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1945.0, "completions/mean_length": 1545.103515625, "completions/mean_terminated_length": 1130.3428955078125, "completions/min_length": 196.0, "completions/min_terminated_length": 196.0, "epoch": 0.10149041873669269, "frac_reward_zero_std": 0.34375, "grad_norm": 2.291518211364746, "learning_rate": 7.159999999999999e-07, "loss": 0.072, "num_tokens": 354822226.0, "reward": 0.39453125, "reward_std": 0.2754097282886505, "rewards/simpleverify_reward/mean": 0.39453125, "rewards/simpleverify_reward/std": 0.4892277717590332, "step": 143, "tools/generated_tokens": 2748.5703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.65625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.37890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1974.0, "completions/mean_length": 1419.529296875, "completions/mean_terminated_length": 1036.5, "completions/min_length": 115.0, "completions/min_terminated_length": 115.0, "epoch": 0.1022001419446416, "frac_reward_zero_std": 0.375, "grad_norm": 4.07804012298584, "learning_rate": 7.14e-07, "loss": 0.0623, "num_tokens": 356893813.0, "reward": 0.42578125, "reward_std": 0.26410993933677673, "rewards/simpleverify_reward/mean": 0.42578125, "rewards/simpleverify_reward/std": 0.4949444830417633, "step": 144, "tools/generated_tokens": 2461.048828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.568359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.322265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2007.0, "completions/mean_length": 1463.287109375, "completions/mean_terminated_length": 1185.5821533203125, "completions/min_length": 394.0, "completions/min_terminated_length": 394.0, "epoch": 0.10290986515259049, "frac_reward_zero_std": 0.28125, "grad_norm": 3.076936721801758, "learning_rate": 7.119999999999999e-07, "loss": 0.0518, "num_tokens": 358982814.0, "reward": 0.458984375, "reward_std": 0.3013579547405243, "rewards/simpleverify_reward/mean": 0.458984375, "rewards/simpleverify_reward/std": 0.49880221486091614, "step": 145, "tools/generated_tokens": 2503.931640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.564453125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.427734375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1978.0, "completions/mean_length": 1467.830078125, "completions/mean_terminated_length": 1037.7030029296875, "completions/min_length": 194.0, "completions/min_terminated_length": 194.0, "epoch": 0.10361958836053939, "frac_reward_zero_std": 0.375, "grad_norm": 3.701575756072998, "learning_rate": 7.1e-07, "loss": 0.0649, "num_tokens": 361218026.0, "reward": 0.396484375, "reward_std": 0.26298588514328003, "rewards/simpleverify_reward/mean": 0.396484375, "rewards/simpleverify_reward/std": 0.4896455705165863, "step": 146, "tools/generated_tokens": 2651.37890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.673828125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.466796875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1962.0, "completions/mean_length": 1615.138671875, "completions/mean_terminated_length": 1240.9853515625, "completions/min_length": 141.0, "completions/min_terminated_length": 141.0, "epoch": 0.10432931156848829, "frac_reward_zero_std": 0.28125, "grad_norm": 1.9344936609268188, "learning_rate": 7.079999999999999e-07, "loss": 0.0819, "num_tokens": 363588702.0, "reward": 0.375, "reward_std": 0.29654520750045776, "rewards/simpleverify_reward/mean": 0.375, "rewards/simpleverify_reward/std": 0.4845963716506958, "step": 147, "tools/generated_tokens": 2874.935546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.69140625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2007.0, "completions/mean_length": 1566.810546875, "completions/mean_terminated_length": 1180.6197509765625, "completions/min_length": 284.0, "completions/min_terminated_length": 284.0, "epoch": 0.1050390347764372, "frac_reward_zero_std": 0.40625, "grad_norm": 1.5118831396102905, "learning_rate": 7.059999999999999e-07, "loss": 0.0528, "num_tokens": 365888878.0, "reward": 0.2890625, "reward_std": 0.22699743509292603, "rewards/simpleverify_reward/mean": 0.2890625, "rewards/simpleverify_reward/std": 0.45377036929130554, "step": 148, "tools/generated_tokens": 2794.88671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.673828125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1995.0, "completions/mean_length": 1569.466796875, "completions/mean_terminated_length": 1060.5040283203125, "completions/min_length": 191.0, "completions/min_terminated_length": 191.0, "epoch": 0.10574875798438609, "frac_reward_zero_std": 0.4375, "grad_norm": 3.080892562866211, "learning_rate": 7.04e-07, "loss": 0.0525, "num_tokens": 368141678.0, "reward": 0.3203125, "reward_std": 0.23488323390483856, "rewards/simpleverify_reward/mean": 0.3203125, "rewards/simpleverify_reward/std": 0.4670529365539551, "step": 149, "tools/generated_tokens": 2722.59765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.638671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.513671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2004.0, "completions/mean_length": 1591.130859375, "completions/mean_terminated_length": 1111.6224365234375, "completions/min_length": 117.0, "completions/min_terminated_length": 117.0, "epoch": 0.10645848119233499, "frac_reward_zero_std": 0.4375, "grad_norm": 1.6319727897644043, "learning_rate": 7.019999999999999e-07, "loss": 0.045, "num_tokens": 370446374.0, "reward": 0.287109375, "reward_std": 0.22765029966831207, "rewards/simpleverify_reward/mean": 0.287109375, "rewards/simpleverify_reward/std": 0.45285552740097046, "step": 150, "tools/generated_tokens": 2800.8984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.572265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1976.0, "completions/mean_length": 1629.306640625, "completions/mean_terminated_length": 1069.136962890625, "completions/min_length": 212.0, "completions/min_terminated_length": 212.0, "epoch": 0.10716820440028389, "frac_reward_zero_std": 0.46875, "grad_norm": 3.067344903945923, "learning_rate": 7e-07, "loss": 0.0295, "num_tokens": 372886774.0, "reward": 0.30078125, "reward_std": 0.19824305176734924, "rewards/simpleverify_reward/mean": 0.30078125, "rewards/simpleverify_reward/std": 0.45904624462127686, "step": 151, "tools/generated_tokens": 2953.828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.5546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1953.0, "completions/mean_length": 1609.5234375, "completions/mean_terminated_length": 1063.350830078125, "completions/min_length": 155.0, "completions/min_terminated_length": 155.0, "epoch": 0.1078779276082328, "frac_reward_zero_std": 0.40625, "grad_norm": 2.407284736633301, "learning_rate": 6.979999999999999e-07, "loss": 0.0563, "num_tokens": 375283875.0, "reward": 0.3046875, "reward_std": 0.2396625280380249, "rewards/simpleverify_reward/mean": 0.3046875, "rewards/simpleverify_reward/std": 0.4607250988483429, "step": 152, "tools/generated_tokens": 2919.720703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.71875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.51171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1963.0, "completions/mean_length": 1631.443359375, "completions/mean_terminated_length": 1200.8441162109375, "completions/min_length": 144.0, "completions/min_terminated_length": 144.0, "epoch": 0.10858765081618169, "frac_reward_zero_std": 0.34375, "grad_norm": 2.747295379638672, "learning_rate": 6.959999999999999e-07, "loss": 0.0587, "num_tokens": 377698618.0, "reward": 0.322265625, "reward_std": 0.2511790692806244, "rewards/simpleverify_reward/mean": 0.322265625, "rewards/simpleverify_reward/std": 0.46780112385749817, "step": 153, "tools/generated_tokens": 2938.68359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.71484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.494140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1993.0, "completions/mean_length": 1596.64453125, "completions/mean_terminated_length": 1156.6099853515625, "completions/min_length": 192.0, "completions/min_terminated_length": 192.0, "epoch": 0.10929737402413059, "frac_reward_zero_std": 0.34375, "grad_norm": 3.36660099029541, "learning_rate": 6.939999999999999e-07, "loss": 0.0438, "num_tokens": 380082559.0, "reward": 0.265625, "reward_std": 0.27108997106552124, "rewards/simpleverify_reward/mean": 0.265625, "rewards/simpleverify_reward/std": 0.44209739565849304, "step": 154, "tools/generated_tokens": 2895.140625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.708984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.00390625 }, { "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": 0.5703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1989.0, "completions/mean_length": 1661.30859375, "completions/mean_terminated_length": 1148.0635986328125, "completions/min_length": 222.0, "completions/min_terminated_length": 222.0, "epoch": 0.11000709723207949, "frac_reward_zero_std": 0.40625, "grad_norm": 5.994703769683838, "learning_rate": 6.919999999999999e-07, "loss": 0.0631, "num_tokens": 382532218.0, "reward": 0.314453125, "reward_std": 0.25703537464141846, "rewards/simpleverify_reward/mean": 0.314453125, "rewards/simpleverify_reward/std": 0.4647517800331116, "step": 155, "tools/generated_tokens": 2981.779296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.73046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.486328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1959.0, "completions/mean_length": 1622.58203125, "completions/mean_terminated_length": 1220.8365478515625, "completions/min_length": 30.0, "completions/min_terminated_length": 30.0, "epoch": 0.11071682044002838, "frac_reward_zero_std": 0.3125, "grad_norm": 6.672191619873047, "learning_rate": 6.9e-07, "loss": 0.0501, "num_tokens": 384969748.0, "reward": 0.38671875, "reward_std": 0.2877660393714905, "rewards/simpleverify_reward/mean": 0.38671875, "rewards/simpleverify_reward/std": 0.48747459053993225, "step": 156, "tools/generated_tokens": 2939.505859375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.71875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.564453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2001.0, "completions/mean_length": 1632.505859375, "completions/mean_terminated_length": 1094.0404052734375, "completions/min_length": 231.0, "completions/min_terminated_length": 231.0, "epoch": 0.11142654364797729, "frac_reward_zero_std": 0.34375, "grad_norm": 1.8129339218139648, "learning_rate": 6.879999999999999e-07, "loss": 0.0255, "num_tokens": 387428110.0, "reward": 0.294921875, "reward_std": 0.27926889061927795, "rewards/simpleverify_reward/mean": 0.294921875, "rewards/simpleverify_reward/std": 0.4564536213874817, "step": 157, "tools/generated_tokens": 2980.298828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7421875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.416015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1950.0, "completions/mean_length": 1517.05859375, "completions/mean_terminated_length": 1138.8294677734375, "completions/min_length": 269.0, "completions/min_terminated_length": 269.0, "epoch": 0.11213626685592619, "frac_reward_zero_std": 0.40625, "grad_norm": 7.0149617195129395, "learning_rate": 6.86e-07, "loss": 0.07, "num_tokens": 389686512.0, "reward": 0.384765625, "reward_std": 0.24847842752933502, "rewards/simpleverify_reward/mean": 0.384765625, "rewards/simpleverify_reward/std": 0.4870156943798065, "step": 158, "tools/generated_tokens": 2730.7578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4609375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1992.0, "completions/mean_length": 1494.451171875, "completions/mean_terminated_length": 1023.1992797851562, "completions/min_length": 123.0, "completions/min_terminated_length": 123.0, "epoch": 0.11284599006387509, "frac_reward_zero_std": 0.28125, "grad_norm": 5.934016227722168, "learning_rate": 6.84e-07, "loss": 0.0471, "num_tokens": 391871144.0, "reward": 0.294921875, "reward_std": 0.30298155546188354, "rewards/simpleverify_reward/mean": 0.294921875, "rewards/simpleverify_reward/std": 0.4564536213874817, "step": 159, "tools/generated_tokens": 2628.7109375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.619140625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.486328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1967.0, "completions/mean_length": 1525.5859375, "completions/mean_terminated_length": 1030.98095703125, "completions/min_length": 75.0, "completions/min_terminated_length": 75.0, "epoch": 0.11355571327182398, "frac_reward_zero_std": 0.3125, "grad_norm": 2.772434949874878, "learning_rate": 6.82e-07, "loss": 0.0738, "num_tokens": 394145352.0, "reward": 0.337890625, "reward_std": 0.28625577688217163, "rewards/simpleverify_reward/mean": 0.337890625, "rewards/simpleverify_reward/std": 0.4734536409378052, "step": 160, "tools/generated_tokens": 2740.5390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.48828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1990.0, "completions/mean_length": 1533.82421875, "completions/mean_terminated_length": 1045.85498046875, "completions/min_length": 111.0, "completions/min_terminated_length": 111.0, "epoch": 0.11426543647977289, "frac_reward_zero_std": 0.28125, "grad_norm": 9.531875610351562, "learning_rate": 6.800000000000001e-07, "loss": 0.0636, "num_tokens": 396433308.0, "reward": 0.314453125, "reward_std": 0.2682219445705414, "rewards/simpleverify_reward/mean": 0.314453125, "rewards/simpleverify_reward/std": 0.4647517800331116, "step": 161, "tools/generated_tokens": 2760.609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.677734375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.54296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1932.0, "completions/mean_length": 1601.998046875, "completions/mean_terminated_length": 1072.132568359375, "completions/min_length": 116.0, "completions/min_terminated_length": 116.0, "epoch": 0.11497515968772179, "frac_reward_zero_std": 0.4375, "grad_norm": 4.991877555847168, "learning_rate": 6.78e-07, "loss": 0.0517, "num_tokens": 398904018.0, "reward": 0.3125, "reward_std": 0.21683457493782043, "rewards/simpleverify_reward/mean": 0.3125, "rewards/simpleverify_reward/std": 0.4639657139778137, "step": 162, "tools/generated_tokens": 2995.62890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.759765625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.00390625 }, { "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": 0.396484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2002.0, "completions/mean_length": 1527.458984375, "completions/mean_terminated_length": 1187.1585693359375, "completions/min_length": 147.0, "completions/min_terminated_length": 147.0, "epoch": 0.1156848828956707, "frac_reward_zero_std": 0.25, "grad_norm": 4.971357345581055, "learning_rate": 6.76e-07, "loss": 0.0433, "num_tokens": 401147660.0, "reward": 0.34375, "reward_std": 0.27940136194229126, "rewards/simpleverify_reward/mean": 0.34375, "rewards/simpleverify_reward/std": 0.4754233956336975, "step": 163, "tools/generated_tokens": 2708.0390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.642578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4921875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1969.0, "completions/mean_length": 1603.21484375, "completions/mean_terminated_length": 1172.1153564453125, "completions/min_length": 16.0, "completions/min_terminated_length": 16.0, "epoch": 0.11639460610361958, "frac_reward_zero_std": 0.40625, "grad_norm": 4.524590492248535, "learning_rate": 6.74e-07, "loss": 0.0359, "num_tokens": 403520408.0, "reward": 0.27734375, "reward_std": 0.23403224349021912, "rewards/simpleverify_reward/mean": 0.27734375, "rewards/simpleverify_reward/std": 0.4481254518032074, "step": 164, "tools/generated_tokens": 2872.150390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.712890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1951.0, "completions/mean_length": 1545.216796875, "completions/mean_terminated_length": 1131.0250244140625, "completions/min_length": 162.0, "completions/min_terminated_length": 162.0, "epoch": 0.11710432931156849, "frac_reward_zero_std": 0.40625, "grad_norm": 3.28016996383667, "learning_rate": 6.72e-07, "loss": 0.0422, "num_tokens": 405844420.0, "reward": 0.2890625, "reward_std": 0.2300632894039154, "rewards/simpleverify_reward/mean": 0.2890625, "rewards/simpleverify_reward/std": 0.45377036929130554, "step": 165, "tools/generated_tokens": 2796.6484375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.681640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.47265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1947.0, "completions/mean_length": 1549.78125, "completions/mean_terminated_length": 1103.2296142578125, "completions/min_length": 198.0, "completions/min_terminated_length": 198.0, "epoch": 0.11781405251951739, "frac_reward_zero_std": 0.53125, "grad_norm": 5.548795700073242, "learning_rate": 6.7e-07, "loss": 0.0496, "num_tokens": 408089641.0, "reward": 0.318359375, "reward_std": 0.19242528080940247, "rewards/simpleverify_reward/mean": 0.318359375, "rewards/simpleverify_reward/std": 0.46629536151885986, "step": 166, "tools/generated_tokens": 2703.935546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.638671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.00390625 }, { "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": 0.40625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1991.0, "completions/mean_length": 1509.7421875, "completions/mean_terminated_length": 1141.4605712890625, "completions/min_length": 166.0, "completions/min_terminated_length": 166.0, "epoch": 0.1185237757274663, "frac_reward_zero_std": 0.4375, "grad_norm": 2.9370882511138916, "learning_rate": 6.68e-07, "loss": 0.0138, "num_tokens": 410317636.0, "reward": 0.40625, "reward_std": 0.21780776977539062, "rewards/simpleverify_reward/mean": 0.40625, "rewards/simpleverify_reward/std": 0.49161264300346375, "step": 167, "tools/generated_tokens": 2692.578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.64453125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.396484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1949.0, "completions/mean_length": 1487.640625, "completions/mean_terminated_length": 1119.5081787109375, "completions/min_length": 142.0, "completions/min_terminated_length": 142.0, "epoch": 0.11923349893541518, "frac_reward_zero_std": 0.28125, "grad_norm": 3.059490203857422, "learning_rate": 6.66e-07, "loss": 0.0392, "num_tokens": 412533017.0, "reward": 0.421875, "reward_std": 0.28113609552383423, "rewards/simpleverify_reward/mean": 0.421875, "rewards/simpleverify_reward/std": 0.49434176087379456, "step": 168, "tools/generated_tokens": 2638.01171875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.62890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.35546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1963.0, "completions/mean_length": 1383.60546875, "completions/mean_terminated_length": 1017.2636108398438, "completions/min_length": 31.0, "completions/min_terminated_length": 31.0, "epoch": 0.11994322214336409, "frac_reward_zero_std": 0.28125, "grad_norm": 5.113927841186523, "learning_rate": 6.64e-07, "loss": 0.0434, "num_tokens": 414618439.0, "reward": 0.37890625, "reward_std": 0.2770920991897583, "rewards/simpleverify_reward/mean": 0.37890625, "rewards/simpleverify_reward/std": 0.4855891764163971, "step": 169, "tools/generated_tokens": 2465.810546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.591796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.41015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1986.0, "completions/mean_length": 1548.751953125, "completions/mean_terminated_length": 1203.039794921875, "completions/min_length": 111.0, "completions/min_terminated_length": 111.0, "epoch": 0.12065294535131299, "frac_reward_zero_std": 0.375, "grad_norm": 5.937819957733154, "learning_rate": 6.62e-07, "loss": 0.0498, "num_tokens": 416871212.0, "reward": 0.3828125, "reward_std": 0.24185073375701904, "rewards/simpleverify_reward/mean": 0.3828125, "rewards/simpleverify_reward/std": 0.486548513174057, "step": 170, "tools/generated_tokens": 2722.033203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.599609375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1990.0, "completions/mean_length": 1679.9296875, "completions/mean_terminated_length": 1128.721923828125, "completions/min_length": 61.0, "completions/min_terminated_length": 61.0, "epoch": 0.1213626685592619, "frac_reward_zero_std": 0.4375, "grad_norm": 4.636204242706299, "learning_rate": 6.6e-07, "loss": 0.051, "num_tokens": 419367994.0, "reward": 0.224609375, "reward_std": 0.2122129201889038, "rewards/simpleverify_reward/mean": 0.224609375, "rewards/simpleverify_reward/std": 0.41773295402526855, "step": 171, "tools/generated_tokens": 3060.166015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.755859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1986.0, "completions/mean_length": 1542.44921875, "completions/mean_terminated_length": 1149.2430419921875, "completions/min_length": 100.0, "completions/min_terminated_length": 100.0, "epoch": 0.12207239176721078, "frac_reward_zero_std": 0.25, "grad_norm": 11.508146286010742, "learning_rate": 6.58e-07, "loss": 0.0918, "num_tokens": 421687168.0, "reward": 0.3984375, "reward_std": 0.3053220808506012, "rewards/simpleverify_reward/mean": 0.3984375, "rewards/simpleverify_reward/std": 0.4900552034378052, "step": 172, "tools/generated_tokens": 2772.490234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.67578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.46875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1975.0, "completions/mean_length": 1607.8671875, "completions/mean_terminated_length": 1220.7684326171875, "completions/min_length": 312.0, "completions/min_terminated_length": 312.0, "epoch": 0.12278211497515969, "frac_reward_zero_std": 0.46875, "grad_norm": 2.223642349243164, "learning_rate": 6.56e-07, "loss": 0.0312, "num_tokens": 424092866.0, "reward": 0.26171875, "reward_std": 0.21688786149024963, "rewards/simpleverify_reward/mean": 0.26171875, "rewards/simpleverify_reward/std": 0.44000017642974854, "step": 173, "tools/generated_tokens": 2922.029296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.720703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.361328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1985.0, "completions/mean_length": 1494.11328125, "completions/mean_terminated_length": 1181.3150634765625, "completions/min_length": 54.0, "completions/min_terminated_length": 54.0, "epoch": 0.12349183818310859, "frac_reward_zero_std": 0.15625, "grad_norm": 10.766304969787598, "learning_rate": 6.54e-07, "loss": 0.0633, "num_tokens": 426291464.0, "reward": 0.431640625, "reward_std": 0.3629853129386902, "rewards/simpleverify_reward/mean": 0.431640625, "rewards/simpleverify_reward/std": 0.4957893490791321, "step": 174, "tools/generated_tokens": 2646.26953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.447265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1994.0, "completions/mean_length": 1529.375, "completions/mean_terminated_length": 1112.395751953125, "completions/min_length": 93.0, "completions/min_terminated_length": 93.0, "epoch": 0.1242015613910575, "frac_reward_zero_std": 0.25, "grad_norm": 14.850948333740234, "learning_rate": 6.52e-07, "loss": 0.0678, "num_tokens": 428613510.0, "reward": 0.296875, "reward_std": 0.2976505160331726, "rewards/simpleverify_reward/mean": 0.296875, "rewards/simpleverify_reward/std": 0.45732781291007996, "step": 175, "tools/generated_tokens": 2801.423828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.69140625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.48828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1966.0, "completions/mean_length": 1585.189453125, "completions/mean_terminated_length": 1143.5762939453125, "completions/min_length": 435.0, "completions/min_terminated_length": 435.0, "epoch": 0.12491128459900638, "frac_reward_zero_std": 0.375, "grad_norm": 6.769768714904785, "learning_rate": 6.5e-07, "loss": 0.0394, "num_tokens": 430947840.0, "reward": 0.287109375, "reward_std": 0.25257647037506104, "rewards/simpleverify_reward/mean": 0.287109375, "rewards/simpleverify_reward/std": 0.45285552740097046, "step": 176, "tools/generated_tokens": 2832.03125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.685546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1987.0, "completions/mean_length": 1549.748046875, "completions/mean_terminated_length": 1149.742919921875, "completions/min_length": 90.0, "completions/min_terminated_length": 90.0, "epoch": 0.1256210078069553, "frac_reward_zero_std": 0.28125, "grad_norm": 7.845015525817871, "learning_rate": 6.48e-07, "loss": 0.059, "num_tokens": 433301427.0, "reward": 0.44921875, "reward_std": 0.28896430134773254, "rewards/simpleverify_reward/mean": 0.44921875, "rewards/simpleverify_reward/std": 0.497901052236557, "step": 177, "tools/generated_tokens": 2860.154296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.71484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1943.0, "completions/mean_length": 1398.810546875, "completions/mean_terminated_length": 1009.296875, "completions/min_length": 196.0, "completions/min_terminated_length": 196.0, "epoch": 0.1263307310149042, "frac_reward_zero_std": 0.40625, "grad_norm": 8.413731575012207, "learning_rate": 6.46e-07, "loss": 0.0439, "num_tokens": 435419964.0, "reward": 0.357421875, "reward_std": 0.23417723178863525, "rewards/simpleverify_reward/mean": 0.357421875, "rewards/simpleverify_reward/std": 0.4797092080116272, "step": 178, "tools/generated_tokens": 2472.63671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.400390625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1951.0, "completions/mean_length": 1472.828125, "completions/mean_terminated_length": 1088.755615234375, "completions/min_length": 101.0, "completions/min_terminated_length": 101.0, "epoch": 0.1270404542228531, "frac_reward_zero_std": 0.375, "grad_norm": 4.50830078125, "learning_rate": 6.44e-07, "loss": 0.0533, "num_tokens": 437524340.0, "reward": 0.40234375, "reward_std": 0.25560417771339417, "rewards/simpleverify_reward/mean": 0.40234375, "rewards/simpleverify_reward/std": 0.4908501207828522, "step": 179, "tools/generated_tokens": 2516.6171875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.59765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2006.0, "completions/mean_length": 1661.283203125, "completions/mean_terminated_length": 1086.83984375, "completions/min_length": 114.0, "completions/min_terminated_length": 114.0, "epoch": 0.127750177430802, "frac_reward_zero_std": 0.40625, "grad_norm": 4.151998043060303, "learning_rate": 6.42e-07, "loss": 0.0507, "num_tokens": 440024128.0, "reward": 0.220703125, "reward_std": 0.2360895723104477, "rewards/simpleverify_reward/mean": 0.220703125, "rewards/simpleverify_reward/std": 0.4151262938976288, "step": 180, "tools/generated_tokens": 3048.55078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.765625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.634765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1967.0, "completions/mean_length": 1730.818359375, "completions/mean_terminated_length": 1181.491943359375, "completions/min_length": 120.0, "completions/min_terminated_length": 120.0, "epoch": 0.12845990063875087, "frac_reward_zero_std": 0.5, "grad_norm": 5.150552272796631, "learning_rate": 6.4e-07, "loss": 0.047, "num_tokens": 442644596.0, "reward": 0.20703125, "reward_std": 0.2021772712469101, "rewards/simpleverify_reward/mean": 0.20703125, "rewards/simpleverify_reward/std": 0.40557438135147095, "step": 181, "tools/generated_tokens": 3224.052734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.822265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4609375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1945.0, "completions/mean_length": 1553.9609375, "completions/mean_terminated_length": 1131.521728515625, "completions/min_length": 253.0, "completions/min_terminated_length": 253.0, "epoch": 0.12916962384669978, "frac_reward_zero_std": 0.21875, "grad_norm": 3.9672884941101074, "learning_rate": 6.38e-07, "loss": 0.0717, "num_tokens": 444994713.0, "reward": 0.3828125, "reward_std": 0.30971473455429077, "rewards/simpleverify_reward/mean": 0.3828125, "rewards/simpleverify_reward/std": 0.486548513174057, "step": 182, "tools/generated_tokens": 2840.677734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.427734375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1983.0, "completions/mean_length": 1515.9921875, "completions/mean_terminated_length": 1119.3992919921875, "completions/min_length": 253.0, "completions/min_terminated_length": 253.0, "epoch": 0.12987934705464868, "frac_reward_zero_std": 0.3125, "grad_norm": 7.301865100860596, "learning_rate": 6.36e-07, "loss": 0.0504, "num_tokens": 447237860.0, "reward": 0.322265625, "reward_std": 0.2862430214881897, "rewards/simpleverify_reward/mean": 0.322265625, "rewards/simpleverify_reward/std": 0.46780112385749817, "step": 183, "tools/generated_tokens": 2713.90234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.65625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.376953125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1965.0, "completions/mean_length": 1478.181640625, "completions/mean_terminated_length": 1133.7930908203125, "completions/min_length": 60.0, "completions/min_terminated_length": 60.0, "epoch": 0.13058907026259758, "frac_reward_zero_std": 0.21875, "grad_norm": 11.162322998046875, "learning_rate": 6.34e-07, "loss": 0.0636, "num_tokens": 449464231.0, "reward": 0.44921875, "reward_std": 0.31167781352996826, "rewards/simpleverify_reward/mean": 0.44921875, "rewards/simpleverify_reward/std": 0.497901052236557, "step": 184, "tools/generated_tokens": 2668.666015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.65625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.447265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1949.0, "completions/mean_length": 1514.171875, "completions/mean_terminated_length": 1087.5335693359375, "completions/min_length": 82.0, "completions/min_terminated_length": 82.0, "epoch": 0.1312987934705465, "frac_reward_zero_std": 0.34375, "grad_norm": 5.840885162353516, "learning_rate": 6.319999999999999e-07, "loss": 0.0653, "num_tokens": 451672873.0, "reward": 0.380859375, "reward_std": 0.28160449862480164, "rewards/simpleverify_reward/mean": 0.380859375, "rewards/simpleverify_reward/std": 0.48607301712036133, "step": 185, "tools/generated_tokens": 2656.2734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.435546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1958.0, "completions/mean_length": 1569.501953125, "completions/mean_terminated_length": 1201.6505126953125, "completions/min_length": 274.0, "completions/min_terminated_length": 274.0, "epoch": 0.1320085166784954, "frac_reward_zero_std": 0.34375, "grad_norm": 4.994154930114746, "learning_rate": 6.3e-07, "loss": 0.0697, "num_tokens": 453978671.0, "reward": 0.365234375, "reward_std": 0.2716749310493469, "rewards/simpleverify_reward/mean": 0.365234375, "rewards/simpleverify_reward/std": 0.4819667339324951, "step": 186, "tools/generated_tokens": 2793.109375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.541015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1938.0, "completions/mean_length": 1664.3203125, "completions/mean_terminated_length": 1212.0679931640625, "completions/min_length": 416.0, "completions/min_terminated_length": 416.0, "epoch": 0.1327182398864443, "frac_reward_zero_std": 0.3125, "grad_norm": 8.782279968261719, "learning_rate": 6.28e-07, "loss": 0.0379, "num_tokens": 456519997.0, "reward": 0.337890625, "reward_std": 0.2817879915237427, "rewards/simpleverify_reward/mean": 0.337890625, "rewards/simpleverify_reward/std": 0.4734536409378052, "step": 187, "tools/generated_tokens": 3088.982421875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.783203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.533203125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1974.0, "completions/mean_length": 1628.95703125, "completions/mean_terminated_length": 1150.30126953125, "completions/min_length": 192.0, "completions/min_terminated_length": 192.0, "epoch": 0.1334279630943932, "frac_reward_zero_std": 0.375, "grad_norm": 5.504629135131836, "learning_rate": 6.26e-07, "loss": 0.0728, "num_tokens": 458934179.0, "reward": 0.33203125, "reward_std": 0.2708049714565277, "rewards/simpleverify_reward/mean": 0.33203125, "rewards/simpleverify_reward/std": 0.47140273451805115, "step": 188, "tools/generated_tokens": 2948.798828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.728515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.322265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1973.0, "completions/mean_length": 1400.9609375, "completions/mean_terminated_length": 1095.573486328125, "completions/min_length": 147.0, "completions/min_terminated_length": 147.0, "epoch": 0.13413768630234207, "frac_reward_zero_std": 0.28125, "grad_norm": 2.778874635696411, "learning_rate": 6.24e-07, "loss": 0.0623, "num_tokens": 460988974.0, "reward": 0.42578125, "reward_std": 0.3061291575431824, "rewards/simpleverify_reward/mean": 0.42578125, "rewards/simpleverify_reward/std": 0.4949444830417633, "step": 189, "tools/generated_tokens": 2452.0390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.36328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2003.0, "completions/mean_length": 1450.529296875, "completions/mean_terminated_length": 1110.1563720703125, "completions/min_length": 36.0, "completions/min_terminated_length": 36.0, "epoch": 0.13484740951029098, "frac_reward_zero_std": 0.3125, "grad_norm": 4.546260833740234, "learning_rate": 6.219999999999999e-07, "loss": 0.0674, "num_tokens": 463158462.0, "reward": 0.421875, "reward_std": 0.2944268584251404, "rewards/simpleverify_reward/mean": 0.421875, "rewards/simpleverify_reward/std": 0.49434176087379456, "step": 190, "tools/generated_tokens": 2585.09375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.423828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2045.0, "completions/mean_length": 1480.8125, "completions/mean_terminated_length": 1063.59326171875, "completions/min_length": 88.0, "completions/min_terminated_length": 88.0, "epoch": 0.13555713271823988, "frac_reward_zero_std": 0.40625, "grad_norm": 3.5124943256378174, "learning_rate": 6.2e-07, "loss": 0.0472, "num_tokens": 465368019.0, "reward": 0.40625, "reward_std": 0.25382810831069946, "rewards/simpleverify_reward/mean": 0.40625, "rewards/simpleverify_reward/std": 0.49161264300346375, "step": 191, "tools/generated_tokens": 2676.767578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.65234375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1981.0, "completions/mean_length": 1495.755859375, "completions/mean_terminated_length": 979.9015502929688, "completions/min_length": 85.0, "completions/min_terminated_length": 85.0, "epoch": 0.13626685592618878, "frac_reward_zero_std": 0.375, "grad_norm": 1.847140908241272, "learning_rate": 6.18e-07, "loss": 0.0558, "num_tokens": 467617191.0, "reward": 0.34765625, "reward_std": 0.2327541708946228, "rewards/simpleverify_reward/mean": 0.34765625, "rewards/simpleverify_reward/std": 0.47669193148612976, "step": 192, "tools/generated_tokens": 2676.638671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.486328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2000.0, "completions/mean_length": 1614.908203125, "completions/mean_terminated_length": 1205.0760498046875, "completions/min_length": 177.0, "completions/min_terminated_length": 177.0, "epoch": 0.1369765791341377, "frac_reward_zero_std": 0.3125, "grad_norm": 16.7745304107666, "learning_rate": 6.16e-07, "loss": 0.039, "num_tokens": 470091071.0, "reward": 0.322265625, "reward_std": 0.2925110459327698, "rewards/simpleverify_reward/mean": 0.322265625, "rewards/simpleverify_reward/std": 0.46780112385749817, "step": 193, "tools/generated_tokens": 3001.9375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.75390625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.443359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1996.0, "completions/mean_length": 1535.4453125, "completions/mean_terminated_length": 1127.2000732421875, "completions/min_length": 34.0, "completions/min_terminated_length": 34.0, "epoch": 0.1376863023420866, "frac_reward_zero_std": 0.375, "grad_norm": 14.338887214660645, "learning_rate": 6.14e-07, "loss": 0.0383, "num_tokens": 472293708.0, "reward": 0.34765625, "reward_std": 0.23600786924362183, "rewards/simpleverify_reward/mean": 0.34765625, "rewards/simpleverify_reward/std": 0.47669193148612976, "step": 194, "tools/generated_tokens": 2644.2421875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.611328125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.3671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1977.0, "completions/mean_length": 1450.037109375, "completions/mean_terminated_length": 1105.69140625, "completions/min_length": 77.0, "completions/min_terminated_length": 77.0, "epoch": 0.1383960255500355, "frac_reward_zero_std": 0.28125, "grad_norm": 14.864490509033203, "learning_rate": 6.119999999999999e-07, "loss": 0.07, "num_tokens": 474392483.0, "reward": 0.392578125, "reward_std": 0.2964054346084595, "rewards/simpleverify_reward/mean": 0.392578125, "rewards/simpleverify_reward/std": 0.4888018071651459, "step": 195, "tools/generated_tokens": 2481.15625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.583984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.494140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1983.0, "completions/mean_length": 1594.650390625, "completions/mean_terminated_length": 1151.8031005859375, "completions/min_length": 36.0, "completions/min_terminated_length": 36.0, "epoch": 0.1391057487579844, "frac_reward_zero_std": 0.375, "grad_norm": 3.5438666343688965, "learning_rate": 6.1e-07, "loss": 0.0545, "num_tokens": 476885804.0, "reward": 0.3203125, "reward_std": 0.27404773235321045, "rewards/simpleverify_reward/mean": 0.3203125, "rewards/simpleverify_reward/std": 0.4670529365539551, "step": 196, "tools/generated_tokens": 3031.009765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.787109375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1949.0, "completions/mean_length": 1521.462890625, "completions/mean_terminated_length": 1085.1893310546875, "completions/min_length": 101.0, "completions/min_terminated_length": 101.0, "epoch": 0.13981547196593327, "frac_reward_zero_std": 0.40625, "grad_norm": 6.371192455291748, "learning_rate": 6.079999999999999e-07, "loss": 0.0324, "num_tokens": 479095770.0, "reward": 0.328125, "reward_std": 0.24057087302207947, "rewards/simpleverify_reward/mean": 0.328125, "rewards/simpleverify_reward/std": 0.4699897766113281, "step": 197, "tools/generated_tokens": 2643.388671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.623046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.52734375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1938.0, "completions/mean_length": 1585.91796875, "completions/mean_terminated_length": 1070.371826171875, "completions/min_length": 56.0, "completions/min_terminated_length": 56.0, "epoch": 0.14052519517388218, "frac_reward_zero_std": 0.40625, "grad_norm": 7.372990131378174, "learning_rate": 6.06e-07, "loss": 0.0501, "num_tokens": 481514629.0, "reward": 0.29296875, "reward_std": 0.2608676254749298, "rewards/simpleverify_reward/mean": 0.29296875, "rewards/simpleverify_reward/std": 0.455569326877594, "step": 198, "tools/generated_tokens": 2917.47265625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.345703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1964.0, "completions/mean_length": 1447.912109375, "completions/mean_terminated_length": 1132.379150390625, "completions/min_length": 163.0, "completions/min_terminated_length": 163.0, "epoch": 0.14123491838183108, "frac_reward_zero_std": 0.21875, "grad_norm": 10.475242614746094, "learning_rate": 6.04e-07, "loss": 0.0401, "num_tokens": 483663068.0, "reward": 0.43359375, "reward_std": 0.31976318359375, "rewards/simpleverify_reward/mean": 0.43359375, "rewards/simpleverify_reward/std": 0.4960552453994751, "step": 199, "tools/generated_tokens": 2567.630859375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.607421875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.44921875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1949.0, "completions/mean_length": 1538.181640625, "completions/mean_terminated_length": 1122.372314453125, "completions/min_length": 101.0, "completions/min_terminated_length": 101.0, "epoch": 0.14194464158977999, "frac_reward_zero_std": 0.34375, "grad_norm": 6.2920613288879395, "learning_rate": 6.019999999999999e-07, "loss": 0.0453, "num_tokens": 485896399.0, "reward": 0.349609375, "reward_std": 0.2658618092536926, "rewards/simpleverify_reward/mean": 0.349609375, "rewards/simpleverify_reward/std": 0.47731292247772217, "step": 200, "tools/generated_tokens": 2702.529296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.638671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.333984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2018.0, "completions/mean_length": 1401.470703125, "completions/mean_terminated_length": 1077.258056640625, "completions/min_length": 107.0, "completions/min_terminated_length": 107.0, "epoch": 0.1426543647977289, "frac_reward_zero_std": 0.375, "grad_norm": 10.922364234924316, "learning_rate": 6e-07, "loss": 0.0611, "num_tokens": 488011521.0, "reward": 0.486328125, "reward_std": 0.25530558824539185, "rewards/simpleverify_reward/mean": 0.486328125, "rewards/simpleverify_reward/std": 0.5003018379211426, "step": 201, "tools/generated_tokens": 2494.720703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.595703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.501953125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2023.0, "completions/mean_length": 1587.84765625, "completions/mean_terminated_length": 1124.0863037109375, "completions/min_length": 45.0, "completions/min_terminated_length": 45.0, "epoch": 0.1433640880056778, "frac_reward_zero_std": 0.34375, "grad_norm": 6.23798942565918, "learning_rate": 5.979999999999999e-07, "loss": 0.0587, "num_tokens": 490342315.0, "reward": 0.3359375, "reward_std": 0.28324151039123535, "rewards/simpleverify_reward/mean": 0.3359375, "rewards/simpleverify_reward/std": 0.4727790653705597, "step": 202, "tools/generated_tokens": 2830.177734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.68359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.37109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1923.0, "completions/mean_length": 1457.2109375, "completions/mean_terminated_length": 1108.6087646484375, "completions/min_length": 139.0, "completions/min_terminated_length": 139.0, "epoch": 0.1440738112136267, "frac_reward_zero_std": 0.21875, "grad_norm": 3.477853775024414, "learning_rate": 5.96e-07, "loss": 0.048, "num_tokens": 492615859.0, "reward": 0.37890625, "reward_std": 0.3094038963317871, "rewards/simpleverify_reward/mean": 0.37890625, "rewards/simpleverify_reward/std": 0.4855891764163971, "step": 203, "tools/generated_tokens": 2715.55078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.681640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.400390625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1943.0, "completions/mean_length": 1487.576171875, "completions/mean_terminated_length": 1113.351806640625, "completions/min_length": 30.0, "completions/min_terminated_length": 30.0, "epoch": 0.1447835344215756, "frac_reward_zero_std": 0.1875, "grad_norm": 6.278757572174072, "learning_rate": 5.939999999999999e-07, "loss": 0.0754, "num_tokens": 494834735.0, "reward": 0.400390625, "reward_std": 0.3189562261104584, "rewards/simpleverify_reward/mean": 0.400390625, "rewards/simpleverify_reward/std": 0.4904567301273346, "step": 204, "tools/generated_tokens": 2649.849609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.642578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.322265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1987.0, "completions/mean_length": 1354.513671875, "completions/mean_terminated_length": 1025.8212890625, "completions/min_length": 88.0, "completions/min_terminated_length": 88.0, "epoch": 0.14549325762952448, "frac_reward_zero_std": 0.1875, "grad_norm": 6.470238208770752, "learning_rate": 5.919999999999999e-07, "loss": 0.066, "num_tokens": 496883404.0, "reward": 0.41015625, "reward_std": 0.3568820357322693, "rewards/simpleverify_reward/mean": 0.41015625, "rewards/simpleverify_reward/std": 0.49234291911125183, "step": 205, "tools/generated_tokens": 2399.193359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.572265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.517578125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1974.0, "completions/mean_length": 1572.3515625, "completions/mean_terminated_length": 1062.04052734375, "completions/min_length": 89.0, "completions/min_terminated_length": 89.0, "epoch": 0.14620298083747338, "frac_reward_zero_std": 0.3125, "grad_norm": 6.1911091804504395, "learning_rate": 5.9e-07, "loss": 0.053, "num_tokens": 499319226.0, "reward": 0.29296875, "reward_std": 0.2767079472541809, "rewards/simpleverify_reward/mean": 0.29296875, "rewards/simpleverify_reward/std": 0.455569326877594, "step": 206, "tools/generated_tokens": 2953.33984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.416015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1928.0, "completions/mean_length": 1457.50390625, "completions/mean_terminated_length": 1039.718994140625, "completions/min_length": 155.0, "completions/min_terminated_length": 155.0, "epoch": 0.14691270404542228, "frac_reward_zero_std": 0.25, "grad_norm": 7.827362537384033, "learning_rate": 5.879999999999999e-07, "loss": 0.0542, "num_tokens": 501569640.0, "reward": 0.375, "reward_std": 0.28827184438705444, "rewards/simpleverify_reward/mean": 0.375, "rewards/simpleverify_reward/std": 0.4845963716506958, "step": 207, "tools/generated_tokens": 2681.966796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.560546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1926.0, "completions/mean_length": 1637.76171875, "completions/mean_terminated_length": 1118.711181640625, "completions/min_length": 39.0, "completions/min_terminated_length": 39.0, "epoch": 0.14762242725337119, "frac_reward_zero_std": 0.34375, "grad_norm": 7.083431243896484, "learning_rate": 5.86e-07, "loss": 0.0317, "num_tokens": 504058023.0, "reward": 0.2734375, "reward_std": 0.2542150020599365, "rewards/simpleverify_reward/mean": 0.2734375, "rewards/simpleverify_reward/std": 0.4461594223976135, "step": 208, "tools/generated_tokens": 3065.4140625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.78125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.439453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1976.0, "completions/mean_length": 1495.806640625, "completions/mean_terminated_length": 1062.9024658203125, "completions/min_length": 57.0, "completions/min_terminated_length": 57.0, "epoch": 0.1483321504613201, "frac_reward_zero_std": 0.28125, "grad_norm": 7.119373321533203, "learning_rate": 5.839999999999999e-07, "loss": 0.0744, "num_tokens": 506331132.0, "reward": 0.3671875, "reward_std": 0.31542953848838806, "rewards/simpleverify_reward/mean": 0.3671875, "rewards/simpleverify_reward/std": 0.48250964283943176, "step": 209, "tools/generated_tokens": 2722.78515625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.41015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1993.0, "completions/mean_length": 1466.51171875, "completions/mean_terminated_length": 1062.16552734375, "completions/min_length": 16.0, "completions/min_terminated_length": 16.0, "epoch": 0.149041873669269, "frac_reward_zero_std": 0.46875, "grad_norm": 2.5552961826324463, "learning_rate": 5.819999999999999e-07, "loss": 0.0465, "num_tokens": 508614390.0, "reward": 0.26953125, "reward_std": 0.20621632039546967, "rewards/simpleverify_reward/mean": 0.26953125, "rewards/simpleverify_reward/std": 0.44415023922920227, "step": 210, "tools/generated_tokens": 2720.525390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.69140625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.458984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2036.0, "completions/mean_length": 1513.84375, "completions/mean_terminated_length": 1060.6787109375, "completions/min_length": 46.0, "completions/min_terminated_length": 46.0, "epoch": 0.1497515968772179, "frac_reward_zero_std": 0.3125, "grad_norm": 6.459166526794434, "learning_rate": 5.8e-07, "loss": 0.0477, "num_tokens": 510868056.0, "reward": 0.291015625, "reward_std": 0.2655189633369446, "rewards/simpleverify_reward/mean": 0.291015625, "rewards/simpleverify_reward/std": 0.45467492938041687, "step": 211, "tools/generated_tokens": 2692.62109375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.65625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.00390625 }, { "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": 0.384765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1904.0, "completions/mean_length": 1500.564453125, "completions/mean_terminated_length": 1158.2000732421875, "completions/min_length": 60.0, "completions/min_terminated_length": 60.0, "epoch": 0.15046132008516677, "frac_reward_zero_std": 0.34375, "grad_norm": 3.529888391494751, "learning_rate": 5.779999999999999e-07, "loss": 0.0416, "num_tokens": 513038229.0, "reward": 0.43359375, "reward_std": 0.2820373475551605, "rewards/simpleverify_reward/mean": 0.43359375, "rewards/simpleverify_reward/std": 0.4960552453994751, "step": 212, "tools/generated_tokens": 2605.3984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.599609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.529296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1981.0, "completions/mean_length": 1597.154296875, "completions/mean_terminated_length": 1090.186767578125, "completions/min_length": 40.0, "completions/min_terminated_length": 40.0, "epoch": 0.15117104329311568, "frac_reward_zero_std": 0.3125, "grad_norm": 5.671728610992432, "learning_rate": 5.76e-07, "loss": 0.0797, "num_tokens": 515536919.0, "reward": 0.310546875, "reward_std": 0.2796097695827484, "rewards/simpleverify_reward/mean": 0.310546875, "rewards/simpleverify_reward/std": 0.46317005157470703, "step": 213, "tools/generated_tokens": 3029.0078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.78125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.419921875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2006.0, "completions/mean_length": 1531.927734375, "completions/mean_terminated_length": 1158.44775390625, "completions/min_length": 51.0, "completions/min_terminated_length": 51.0, "epoch": 0.15188076650106458, "frac_reward_zero_std": 0.28125, "grad_norm": 4.772290229797363, "learning_rate": 5.739999999999999e-07, "loss": 0.0329, "num_tokens": 517898936.0, "reward": 0.263671875, "reward_std": 0.2906274199485779, "rewards/simpleverify_reward/mean": 0.263671875, "rewards/simpleverify_reward/std": 0.4410543739795685, "step": 214, "tools/generated_tokens": 2865.318359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.724609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.380859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1996.0, "completions/mean_length": 1423.791015625, "completions/mean_terminated_length": 1044.6845703125, "completions/min_length": 77.0, "completions/min_terminated_length": 77.0, "epoch": 0.15259048970901348, "frac_reward_zero_std": 0.28125, "grad_norm": 2.0860750675201416, "learning_rate": 5.719999999999999e-07, "loss": 0.0491, "num_tokens": 519998620.0, "reward": 0.421875, "reward_std": 0.2823367118835449, "rewards/simpleverify_reward/mean": 0.421875, "rewards/simpleverify_reward/std": 0.49434176087379456, "step": 215, "tools/generated_tokens": 2514.138671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.59375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.45703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1898.0, "completions/mean_length": 1500.794921875, "completions/mean_terminated_length": 1040.1978759765625, "completions/min_length": 55.0, "completions/min_terminated_length": 55.0, "epoch": 0.1533002129169624, "frac_reward_zero_std": 0.28125, "grad_norm": 2.775409460067749, "learning_rate": 5.699999999999999e-07, "loss": 0.0526, "num_tokens": 522258813.0, "reward": 0.33984375, "reward_std": 0.27700862288475037, "rewards/simpleverify_reward/mean": 0.33984375, "rewards/simpleverify_reward/std": 0.4741191864013672, "step": 216, "tools/generated_tokens": 2721.69140625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.54296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1966.0, "completions/mean_length": 1611.05859375, "completions/mean_terminated_length": 1091.957275390625, "completions/min_length": 145.0, "completions/min_terminated_length": 145.0, "epoch": 0.1540099361249113, "frac_reward_zero_std": 0.21875, "grad_norm": 2.2952756881713867, "learning_rate": 5.679999999999999e-07, "loss": 0.0596, "num_tokens": 524713898.0, "reward": 0.30078125, "reward_std": 0.30984920263290405, "rewards/simpleverify_reward/mean": 0.30078125, "rewards/simpleverify_reward/std": 0.45904624462127686, "step": 217, "tools/generated_tokens": 2973.90234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.75, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1970.0, "completions/mean_length": 1397.166015625, "completions/mean_terminated_length": 1105.5142822265625, "completions/min_length": 119.0, "completions/min_terminated_length": 119.0, "epoch": 0.1547196593328602, "frac_reward_zero_std": 0.09375, "grad_norm": 5.298105716705322, "learning_rate": 5.66e-07, "loss": 0.0742, "num_tokens": 526832790.0, "reward": 0.533203125, "reward_std": 0.39117127656936646, "rewards/simpleverify_reward/mean": 0.533203125, "rewards/simpleverify_reward/std": 0.4993842542171478, "step": 218, "tools/generated_tokens": 2523.16015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.611328125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.330078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2004.0, "completions/mean_length": 1394.18359375, "completions/mean_terminated_length": 1072.0408935546875, "completions/min_length": 29.0, "completions/min_terminated_length": 29.0, "epoch": 0.1554293825408091, "frac_reward_zero_std": 0.15625, "grad_norm": 3.0254647731781006, "learning_rate": 5.639999999999999e-07, "loss": 0.054, "num_tokens": 529016642.0, "reward": 0.462890625, "reward_std": 0.34186914563179016, "rewards/simpleverify_reward/mean": 0.462890625, "rewards/simpleverify_reward/std": 0.4991086423397064, "step": 219, "tools/generated_tokens": 2599.150390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.654296875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.37109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1972.0, "completions/mean_length": 1479.173828125, "completions/mean_terminated_length": 1144.6397705078125, "completions/min_length": 40.0, "completions/min_terminated_length": 40.0, "epoch": 0.15613910574875797, "frac_reward_zero_std": 0.28125, "grad_norm": 4.364960670471191, "learning_rate": 5.620000000000001e-07, "loss": 0.0616, "num_tokens": 531210868.0, "reward": 0.478515625, "reward_std": 0.2785497307777405, "rewards/simpleverify_reward/mean": 0.478515625, "rewards/simpleverify_reward/std": 0.5000267624855042, "step": 220, "tools/generated_tokens": 2620.142578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1939.0, "completions/mean_length": 1516.86328125, "completions/mean_terminated_length": 1165.4610595703125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.15684882895670688, "frac_reward_zero_std": 0.28125, "grad_norm": 2.6563122272491455, "learning_rate": 5.6e-07, "loss": 0.0528, "num_tokens": 533345133.0, "reward": 0.388671875, "reward_std": 0.2821839153766632, "rewards/simpleverify_reward/mean": 0.388671875, "rewards/simpleverify_reward/std": 0.4879252314567566, "step": 221, "tools/generated_tokens": 2559.126953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.439453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1983.0, "completions/mean_length": 1553.041015625, "completions/mean_terminated_length": 1169.243896484375, "completions/min_length": 179.0, "completions/min_terminated_length": 179.0, "epoch": 0.15755855216465578, "frac_reward_zero_std": 0.28125, "grad_norm": 6.605661869049072, "learning_rate": 5.58e-07, "loss": 0.0603, "num_tokens": 535686451.0, "reward": 0.330078125, "reward_std": 0.3082415461540222, "rewards/simpleverify_reward/mean": 0.330078125, "rewards/simpleverify_reward/std": 0.47070086002349854, "step": 222, "tools/generated_tokens": 2829.75, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.701171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.376953125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1981.0, "completions/mean_length": 1416.216796875, "completions/mean_terminated_length": 1033.97802734375, "completions/min_length": 69.0, "completions/min_terminated_length": 69.0, "epoch": 0.15826827537260468, "frac_reward_zero_std": 0.25, "grad_norm": 3.461460590362549, "learning_rate": 5.560000000000001e-07, "loss": 0.0859, "num_tokens": 537812886.0, "reward": 0.361328125, "reward_std": 0.33688637614250183, "rewards/simpleverify_reward/mean": 0.361328125, "rewards/simpleverify_reward/std": 0.48085519671440125, "step": 223, "tools/generated_tokens": 2524.583984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.607421875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.48828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1940.0, "completions/mean_length": 1553.373046875, "completions/mean_terminated_length": 1083.175537109375, "completions/min_length": 29.0, "completions/min_terminated_length": 29.0, "epoch": 0.1589779985805536, "frac_reward_zero_std": 0.34375, "grad_norm": 4.680582523345947, "learning_rate": 5.54e-07, "loss": 0.0639, "num_tokens": 540188511.0, "reward": 0.349609375, "reward_std": 0.2720368802547455, "rewards/simpleverify_reward/mean": 0.349609375, "rewards/simpleverify_reward/std": 0.47731292247772217, "step": 224, "tools/generated_tokens": 2872.685546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.720703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.349609375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1942.0, "completions/mean_length": 1450.791015625, "completions/mean_terminated_length": 1133.2733154296875, "completions/min_length": 48.0, "completions/min_terminated_length": 48.0, "epoch": 0.1596877217885025, "frac_reward_zero_std": 0.28125, "grad_norm": 1.8984034061431885, "learning_rate": 5.520000000000001e-07, "loss": 0.0581, "num_tokens": 542258312.0, "reward": 0.431640625, "reward_std": 0.30153536796569824, "rewards/simpleverify_reward/mean": 0.431640625, "rewards/simpleverify_reward/std": 0.4957893490791321, "step": 225, "tools/generated_tokens": 2476.62109375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.560546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.439453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2007.0, "completions/mean_length": 1598.390625, "completions/mean_terminated_length": 1246.3275146484375, "completions/min_length": 111.0, "completions/min_terminated_length": 111.0, "epoch": 0.1603974449964514, "frac_reward_zero_std": 0.25, "grad_norm": 1.9901866912841797, "learning_rate": 5.5e-07, "loss": 0.0555, "num_tokens": 544569523.0, "reward": 0.35546875, "reward_std": 0.2946563959121704, "rewards/simpleverify_reward/mean": 0.35546875, "rewards/simpleverify_reward/std": 0.47912323474884033, "step": 226, "tools/generated_tokens": 2819.693359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.666015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.29296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1952.0, "completions/mean_length": 1336.12109375, "completions/mean_terminated_length": 1042.7845458984375, "completions/min_length": 160.0, "completions/min_terminated_length": 160.0, "epoch": 0.1611071682044003, "frac_reward_zero_std": 0.34375, "grad_norm": 1.7495880126953125, "learning_rate": 5.48e-07, "loss": 0.0598, "num_tokens": 546546809.0, "reward": 0.419921875, "reward_std": 0.2559572458267212, "rewards/simpleverify_reward/mean": 0.419921875, "rewards/simpleverify_reward/std": 0.4940285086631775, "step": 227, "tools/generated_tokens": 2313.85546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.537109375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.416015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1943.0, "completions/mean_length": 1477.1953125, "completions/mean_terminated_length": 1072.04345703125, "completions/min_length": 145.0, "completions/min_terminated_length": 145.0, "epoch": 0.16181689141234917, "frac_reward_zero_std": 0.34375, "grad_norm": 2.791768789291382, "learning_rate": 5.46e-07, "loss": 0.0382, "num_tokens": 548728219.0, "reward": 0.375, "reward_std": 0.2895514965057373, "rewards/simpleverify_reward/mean": 0.375, "rewards/simpleverify_reward/std": 0.4845963716506958, "step": 228, "tools/generated_tokens": 2597.833984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.615234375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.40234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1996.0, "completions/mean_length": 1494.310546875, "completions/mean_terminated_length": 1123.761474609375, "completions/min_length": 55.0, "completions/min_terminated_length": 55.0, "epoch": 0.16252661462029808, "frac_reward_zero_std": 0.25, "grad_norm": 2.5284523963928223, "learning_rate": 5.44e-07, "loss": 0.065, "num_tokens": 550974046.0, "reward": 0.42578125, "reward_std": 0.2920331656932831, "rewards/simpleverify_reward/mean": 0.42578125, "rewards/simpleverify_reward/std": 0.4949444830417633, "step": 229, "tools/generated_tokens": 2695.4921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.37890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1951.0, "completions/mean_length": 1366.326171875, "completions/mean_terminated_length": 950.4622192382812, "completions/min_length": 32.0, "completions/min_terminated_length": 32.0, "epoch": 0.16323633782824698, "frac_reward_zero_std": 0.34375, "grad_norm": 3.5371363162994385, "learning_rate": 5.420000000000001e-07, "loss": 0.0693, "num_tokens": 553022067.0, "reward": 0.423828125, "reward_std": 0.2545926570892334, "rewards/simpleverify_reward/mean": 0.423828125, "rewards/simpleverify_reward/std": 0.4946470856666565, "step": 230, "tools/generated_tokens": 2386.556640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.576171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.513671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1989.0, "completions/mean_length": 1628.0703125, "completions/mean_terminated_length": 1184.530029296875, "completions/min_length": 103.0, "completions/min_terminated_length": 103.0, "epoch": 0.16394606103619588, "frac_reward_zero_std": 0.21875, "grad_norm": 2.9785397052764893, "learning_rate": 5.4e-07, "loss": 0.0702, "num_tokens": 555498640.0, "reward": 0.322265625, "reward_std": 0.2969973683357239, "rewards/simpleverify_reward/mean": 0.322265625, "rewards/simpleverify_reward/std": 0.46780112385749817, "step": 231, "tools/generated_tokens": 3039.609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.771484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.388671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1964.0, "completions/mean_length": 1404.294921875, "completions/mean_terminated_length": 995.038330078125, "completions/min_length": 97.0, "completions/min_terminated_length": 97.0, "epoch": 0.1646557842441448, "frac_reward_zero_std": 0.34375, "grad_norm": 2.195366382598877, "learning_rate": 5.38e-07, "loss": 0.029, "num_tokens": 557599085.0, "reward": 0.3671875, "reward_std": 0.25050824880599976, "rewards/simpleverify_reward/mean": 0.3671875, "rewards/simpleverify_reward/std": 0.48250964283943176, "step": 232, "tools/generated_tokens": 2460.443359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.58984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.470703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1977.0, "completions/mean_length": 1547.9296875, "completions/mean_terminated_length": 1106.261962890625, "completions/min_length": 160.0, "completions/min_terminated_length": 160.0, "epoch": 0.1653655074520937, "frac_reward_zero_std": 0.34375, "grad_norm": 2.062624931335449, "learning_rate": 5.36e-07, "loss": 0.0552, "num_tokens": 559880082.0, "reward": 0.31640625, "reward_std": 0.2790224552154541, "rewards/simpleverify_reward/mean": 0.31640625, "rewards/simpleverify_reward/std": 0.46552830934524536, "step": 233, "tools/generated_tokens": 2743.169921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.302734375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1977.0, "completions/mean_length": 1389.943359375, "completions/mean_terminated_length": 1105.761962890625, "completions/min_length": 218.0, "completions/min_terminated_length": 218.0, "epoch": 0.1660752306600426, "frac_reward_zero_std": 0.21875, "grad_norm": 5.985572338104248, "learning_rate": 5.34e-07, "loss": 0.042, "num_tokens": 561955477.0, "reward": 0.46875, "reward_std": 0.29918256402015686, "rewards/simpleverify_reward/mean": 0.46875, "rewards/simpleverify_reward/std": 0.4995105266571045, "step": 234, "tools/generated_tokens": 2467.005859375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.587890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.44140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2004.0, "completions/mean_length": 1487.265625, "completions/mean_terminated_length": 1044.1678466796875, "completions/min_length": 128.0, "completions/min_terminated_length": 128.0, "epoch": 0.16678495386799147, "frac_reward_zero_std": 0.3125, "grad_norm": 2.4282681941986084, "learning_rate": 5.32e-07, "loss": 0.0866, "num_tokens": 564255080.0, "reward": 0.34765625, "reward_std": 0.2714400887489319, "rewards/simpleverify_reward/mean": 0.34765625, "rewards/simpleverify_reward/std": 0.47669193148612976, "step": 235, "tools/generated_tokens": 2755.728515625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.69140625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.337890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1977.0, "completions/mean_length": 1406.9296875, "completions/mean_terminated_length": 1080.7020263671875, "completions/min_length": 44.0, "completions/min_terminated_length": 44.0, "epoch": 0.16749467707594037, "frac_reward_zero_std": 0.28125, "grad_norm": 8.07703971862793, "learning_rate": 5.3e-07, "loss": 0.0388, "num_tokens": 566465779.0, "reward": 0.39453125, "reward_std": 0.32081496715545654, "rewards/simpleverify_reward/mean": 0.39453125, "rewards/simpleverify_reward/std": 0.4892277717590332, "step": 236, "tools/generated_tokens": 2632.98046875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.37890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1970.0, "completions/mean_length": 1440.736328125, "completions/mean_terminated_length": 1070.817626953125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.16820440028388928, "frac_reward_zero_std": 0.1875, "grad_norm": 6.022457122802734, "learning_rate": 5.28e-07, "loss": 0.0868, "num_tokens": 568678300.0, "reward": 0.4453125, "reward_std": 0.33135366439819336, "rewards/simpleverify_reward/mean": 0.4453125, "rewards/simpleverify_reward/std": 0.49748632311820984, "step": 237, "tools/generated_tokens": 2627.734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.65625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.439453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1950.0, "completions/mean_length": 1452.685546875, "completions/mean_terminated_length": 987.2578735351562, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.16891412349183818, "frac_reward_zero_std": 0.25, "grad_norm": 7.14925479888916, "learning_rate": 5.26e-07, "loss": 0.063, "num_tokens": 570880114.0, "reward": 0.318359375, "reward_std": 0.28388869762420654, "rewards/simpleverify_reward/mean": 0.318359375, "rewards/simpleverify_reward/std": 0.46629536151885986, "step": 238, "tools/generated_tokens": 2624.580078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.64453125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.33984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1970.0, "completions/mean_length": 1414.5859375, "completions/mean_terminated_length": 1089.1153564453125, "completions/min_length": 89.0, "completions/min_terminated_length": 89.0, "epoch": 0.16962384669978708, "frac_reward_zero_std": 0.1875, "grad_norm": 9.01520824432373, "learning_rate": 5.24e-07, "loss": 0.0535, "num_tokens": 572927625.0, "reward": 0.474609375, "reward_std": 0.3374847173690796, "rewards/simpleverify_reward/mean": 0.474609375, "rewards/simpleverify_reward/std": 0.4998432695865631, "step": 239, "tools/generated_tokens": 2412.134765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.556640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.3203125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1955.0, "completions/mean_length": 1417.154296875, "completions/mean_terminated_length": 1122.27294921875, "completions/min_length": 151.0, "completions/min_terminated_length": 151.0, "epoch": 0.170333569907736, "frac_reward_zero_std": 0.28125, "grad_norm": 3.5529870986938477, "learning_rate": 5.22e-07, "loss": 0.0586, "num_tokens": 575028917.0, "reward": 0.48828125, "reward_std": 0.27378636598587036, "rewards/simpleverify_reward/mean": 0.48828125, "rewards/simpleverify_reward/std": 0.5003514885902405, "step": 240, "tools/generated_tokens": 2480.953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.583984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.435546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1979.0, "completions/mean_length": 1474.0859375, "completions/mean_terminated_length": 1031.23876953125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1710432931156849, "frac_reward_zero_std": 0.25, "grad_norm": 7.057657241821289, "learning_rate": 5.2e-07, "loss": 0.0659, "num_tokens": 577305201.0, "reward": 0.37890625, "reward_std": 0.29782232642173767, "rewards/simpleverify_reward/mean": 0.37890625, "rewards/simpleverify_reward/std": 0.4855891764163971, "step": 241, "tools/generated_tokens": 2705.591796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.68359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1959.0, "completions/mean_length": 1373.134765625, "completions/mean_terminated_length": 1049.590087890625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1717530163236338, "frac_reward_zero_std": 0.25, "grad_norm": 3.5622315406799316, "learning_rate": 5.18e-07, "loss": 0.0782, "num_tokens": 579223118.0, "reward": 0.44921875, "reward_std": 0.29603248834609985, "rewards/simpleverify_reward/mean": 0.44921875, "rewards/simpleverify_reward/std": 0.497901052236557, "step": 242, "tools/generated_tokens": 2259.41015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.48828125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2036.0, "completions/mean_length": 1509.125, "completions/mean_terminated_length": 1116.5810546875, "completions/min_length": 59.0, "completions/min_terminated_length": 59.0, "epoch": 0.17246273953158267, "frac_reward_zero_std": 0.34375, "grad_norm": 2.142519950866699, "learning_rate": 5.16e-07, "loss": 0.0895, "num_tokens": 581550894.0, "reward": 0.3671875, "reward_std": 0.2864014208316803, "rewards/simpleverify_reward/mean": 0.3671875, "rewards/simpleverify_reward/std": 0.48250964283943176, "step": 243, "tools/generated_tokens": 2821.033203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7109375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.400390625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2000.0, "completions/mean_length": 1437.755859375, "completions/mean_terminated_length": 1030.6351318359375, "completions/min_length": 71.0, "completions/min_terminated_length": 71.0, "epoch": 0.17317246273953157, "frac_reward_zero_std": 0.21875, "grad_norm": 3.2735607624053955, "learning_rate": 5.14e-07, "loss": 0.0653, "num_tokens": 583773619.0, "reward": 0.333984375, "reward_std": 0.3170202672481537, "rewards/simpleverify_reward/mean": 0.333984375, "rewards/simpleverify_reward/std": 0.47209542989730835, "step": 244, "tools/generated_tokens": 2633.74609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.662109375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.505859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1987.0, "completions/mean_length": 1570.64453125, "completions/mean_terminated_length": 1082.7431640625, "completions/min_length": 111.0, "completions/min_terminated_length": 111.0, "epoch": 0.17388218594748048, "frac_reward_zero_std": 0.40625, "grad_norm": 5.424414157867432, "learning_rate": 5.12e-07, "loss": 0.0462, "num_tokens": 586145599.0, "reward": 0.25, "reward_std": 0.24866777658462524, "rewards/simpleverify_reward/mean": 0.25, "rewards/simpleverify_reward/std": 0.43343618512153625, "step": 245, "tools/generated_tokens": 2846.68359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.712890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.451171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1954.0, "completions/mean_length": 1451.46484375, "completions/mean_terminated_length": 961.6441040039062, "completions/min_length": 1.0, "completions/min_terminated_length": 1.0, "epoch": 0.17459190915542938, "frac_reward_zero_std": 0.46875, "grad_norm": 2.491034507751465, "learning_rate": 5.1e-07, "loss": 0.0229, "num_tokens": 588327452.0, "reward": 0.3671875, "reward_std": 0.20950090885162354, "rewards/simpleverify_reward/mean": 0.3671875, "rewards/simpleverify_reward/std": 0.48250964283943176, "step": 246, "tools/generated_tokens": 2589.169921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.638671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.203125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1911.0, "completions/mean_length": 1220.490234375, "completions/mean_terminated_length": 1010.7622680664062, "completions/min_length": 68.0, "completions/min_terminated_length": 68.0, "epoch": 0.17530163236337828, "frac_reward_zero_std": 0.25, "grad_norm": 1.800126314163208, "learning_rate": 5.079999999999999e-07, "loss": 0.1084, "num_tokens": 590080404.0, "reward": 0.4921875, "reward_std": 0.3160877823829651, "rewards/simpleverify_reward/mean": 0.4921875, "rewards/simpleverify_reward/std": 0.5004279017448425, "step": 247, "tools/generated_tokens": 2002.888671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.4296875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1997.0, "completions/mean_length": 1337.37109375, "completions/mean_terminated_length": 1038.566650390625, "completions/min_length": 27.0, "completions/min_terminated_length": 27.0, "epoch": 0.1760113555713272, "frac_reward_zero_std": 0.1875, "grad_norm": 2.6611547470092773, "learning_rate": 5.06e-07, "loss": 0.0485, "num_tokens": 591998924.0, "reward": 0.46484375, "reward_std": 0.31165874004364014, "rewards/simpleverify_reward/mean": 0.46484375, "rewards/simpleverify_reward/std": 0.49925029277801514, "step": 248, "tools/generated_tokens": 2268.1328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.51171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.41015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1899.0, "completions/mean_length": 1427.318359375, "completions/mean_terminated_length": 996.4138793945312, "completions/min_length": 94.0, "completions/min_terminated_length": 94.0, "epoch": 0.1767210787792761, "frac_reward_zero_std": 0.25, "grad_norm": 2.5148565769195557, "learning_rate": 5.04e-07, "loss": 0.0573, "num_tokens": 594118906.0, "reward": 0.427734375, "reward_std": 0.3192230463027954, "rewards/simpleverify_reward/mean": 0.427734375, "rewards/simpleverify_reward/std": 0.4952339828014374, "step": 249, "tools/generated_tokens": 2516.94140625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.33984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1994.0, "completions/mean_length": 1390.615234375, "completions/mean_terminated_length": 1054.2574462890625, "completions/min_length": 42.0, "completions/min_terminated_length": 42.0, "epoch": 0.177430801987225, "frac_reward_zero_std": 0.28125, "grad_norm": 5.244366645812988, "learning_rate": 5.02e-07, "loss": 0.0566, "num_tokens": 596311603.0, "reward": 0.4140625, "reward_std": 0.3054685592651367, "rewards/simpleverify_reward/mean": 0.4140625, "rewards/simpleverify_reward/std": 0.49304109811782837, "step": 250, "tools/generated_tokens": 2596.943359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.310546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1950.0, "completions/mean_length": 1381.8359375, "completions/mean_terminated_length": 1083.7138671875, "completions/min_length": 46.0, "completions/min_terminated_length": 46.0, "epoch": 0.17814052519517387, "frac_reward_zero_std": 0.40625, "grad_norm": 0.8694642782211304, "learning_rate": 5e-07, "loss": 0.0186, "num_tokens": 598225944.0, "reward": 0.46484375, "reward_std": 0.25274285674095154, "rewards/simpleverify_reward/mean": 0.46484375, "rewards/simpleverify_reward/std": 0.49925029277801514, "step": 251, "tools/generated_tokens": 2254.001953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.478515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1955.0, "completions/mean_length": 1422.01953125, "completions/mean_terminated_length": 1138.4432373046875, "completions/min_length": 171.0, "completions/min_terminated_length": 171.0, "epoch": 0.17885024840312277, "frac_reward_zero_std": 0.25, "grad_norm": 2.417189359664917, "learning_rate": 4.979999999999999e-07, "loss": 0.0472, "num_tokens": 600391117.0, "reward": 0.39453125, "reward_std": 0.29213133454322815, "rewards/simpleverify_reward/mean": 0.39453125, "rewards/simpleverify_reward/std": 0.4892277717590332, "step": 252, "tools/generated_tokens": 2591.921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6328125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.30859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1998.0, "completions/mean_length": 1386.57421875, "completions/mean_terminated_length": 1092.31640625, "completions/min_length": 65.0, "completions/min_terminated_length": 65.0, "epoch": 0.17955997161107168, "frac_reward_zero_std": 0.1875, "grad_norm": 2.7489123344421387, "learning_rate": 4.96e-07, "loss": 0.0607, "num_tokens": 602471032.0, "reward": 0.376953125, "reward_std": 0.3161897659301758, "rewards/simpleverify_reward/mean": 0.376953125, "rewards/simpleverify_reward/std": 0.4850969910621643, "step": 253, "tools/generated_tokens": 2473.29296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.595703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.380859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1991.0, "completions/mean_length": 1468.791015625, "completions/mean_terminated_length": 1113.6309814453125, "completions/min_length": 31.0, "completions/min_terminated_length": 31.0, "epoch": 0.18026969481902058, "frac_reward_zero_std": 0.28125, "grad_norm": 2.0590574741363525, "learning_rate": 4.94e-07, "loss": 0.0345, "num_tokens": 604684191.0, "reward": 0.361328125, "reward_std": 0.2838160991668701, "rewards/simpleverify_reward/mean": 0.361328125, "rewards/simpleverify_reward/std": 0.48085519671440125, "step": 254, "tools/generated_tokens": 2665.20703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.650390625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.439453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1985.0, "completions/mean_length": 1471.8203125, "completions/mean_terminated_length": 1023.5609741210938, "completions/min_length": 64.0, "completions/min_terminated_length": 64.0, "epoch": 0.18097941802696949, "frac_reward_zero_std": 0.40625, "grad_norm": 2.588332176208496, "learning_rate": 4.92e-07, "loss": 0.0401, "num_tokens": 606965445.0, "reward": 0.28515625, "reward_std": 0.22578006982803345, "rewards/simpleverify_reward/mean": 0.28515625, "rewards/simpleverify_reward/std": 0.45193037390708923, "step": 255, "tools/generated_tokens": 2713.44921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.4375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1954.0, "completions/mean_length": 1546.458984375, "completions/mean_terminated_length": 1156.80908203125, "completions/min_length": 41.0, "completions/min_terminated_length": 41.0, "epoch": 0.1816891412349184, "frac_reward_zero_std": 0.34375, "grad_norm": 5.781148910522461, "learning_rate": 4.9e-07, "loss": 0.0661, "num_tokens": 609262298.0, "reward": 0.36328125, "reward_std": 0.2743372321128845, "rewards/simpleverify_reward/mean": 0.36328125, "rewards/simpleverify_reward/std": 0.4814152419567108, "step": 256, "tools/generated_tokens": 2755.056640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.666015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1958.0, "completions/mean_length": 1582.353515625, "completions/mean_terminated_length": 1161.9925537109375, "completions/min_length": 60.0, "completions/min_terminated_length": 60.0, "epoch": 0.1823988644428673, "frac_reward_zero_std": 0.375, "grad_norm": 4.687938690185547, "learning_rate": 4.879999999999999e-07, "loss": 0.061, "num_tokens": 611632307.0, "reward": 0.314453125, "reward_std": 0.23828986287117004, "rewards/simpleverify_reward/mean": 0.314453125, "rewards/simpleverify_reward/std": 0.4647517800331116, "step": 257, "tools/generated_tokens": 2833.173828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.689453125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2019.0, "completions/mean_length": 1439.63671875, "completions/mean_terminated_length": 1098.3597412109375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.18310858765081617, "frac_reward_zero_std": 0.3125, "grad_norm": 4.834598064422607, "learning_rate": 4.86e-07, "loss": 0.0683, "num_tokens": 613797642.0, "reward": 0.435546875, "reward_std": 0.2888004183769226, "rewards/simpleverify_reward/mean": 0.435546875, "rewards/simpleverify_reward/std": 0.49631330370903015, "step": 258, "tools/generated_tokens": 2575.638671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.615234375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.451171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1952.0, "completions/mean_length": 1452.138671875, "completions/mean_terminated_length": 963.4768676757812, "completions/min_length": 57.0, "completions/min_terminated_length": 57.0, "epoch": 0.18381831085876507, "frac_reward_zero_std": 0.46875, "grad_norm": 1.9257456064224243, "learning_rate": 4.839999999999999e-07, "loss": 0.0453, "num_tokens": 616135209.0, "reward": 0.310546875, "reward_std": 0.22058087587356567, "rewards/simpleverify_reward/mean": 0.310546875, "rewards/simpleverify_reward/std": 0.46317005157470703, "step": 259, "tools/generated_tokens": 2759.673828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.716796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.357421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1990.0, "completions/mean_length": 1424.478515625, "completions/mean_terminated_length": 1077.9757080078125, "completions/min_length": 49.0, "completions/min_terminated_length": 49.0, "epoch": 0.18452803406671398, "frac_reward_zero_std": 0.21875, "grad_norm": 5.305910587310791, "learning_rate": 4.82e-07, "loss": 0.0719, "num_tokens": 618225340.0, "reward": 0.5, "reward_std": 0.3109908699989319, "rewards/simpleverify_reward/mean": 0.5, "rewards/simpleverify_reward/std": 0.5004889965057373, "step": 260, "tools/generated_tokens": 2484.62890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.435546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1984.0, "completions/mean_length": 1499.685546875, "completions/mean_terminated_length": 1076.5916748046875, "completions/min_length": 95.0, "completions/min_terminated_length": 95.0, "epoch": 0.18523775727466288, "frac_reward_zero_std": 0.3125, "grad_norm": 2.9545371532440186, "learning_rate": 4.8e-07, "loss": 0.0741, "num_tokens": 620527907.0, "reward": 0.36328125, "reward_std": 0.2822868824005127, "rewards/simpleverify_reward/mean": 0.36328125, "rewards/simpleverify_reward/std": 0.4814152419567108, "step": 261, "tools/generated_tokens": 2780.076171875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.69921875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.298828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1988.0, "completions/mean_length": 1322.4921875, "completions/mean_terminated_length": 1013.4540405273438, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.18594748048261178, "frac_reward_zero_std": 0.40625, "grad_norm": 6.618563175201416, "learning_rate": 4.779999999999999e-07, "loss": 0.0552, "num_tokens": 622460913.0, "reward": 0.443359375, "reward_std": 0.24856244027614594, "rewards/simpleverify_reward/mean": 0.443359375, "rewards/simpleverify_reward/std": 0.49726733565330505, "step": 262, "tools/generated_tokens": 2265.845703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.513671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1881.0, "completions/mean_length": 1363.90234375, "completions/mean_terminated_length": 966.956787109375, "completions/min_length": 114.0, "completions/min_terminated_length": 114.0, "epoch": 0.18665720369056069, "frac_reward_zero_std": 0.21875, "grad_norm": 3.436565399169922, "learning_rate": 4.76e-07, "loss": 0.0613, "num_tokens": 624486546.0, "reward": 0.34765625, "reward_std": 0.30792373418807983, "rewards/simpleverify_reward/mean": 0.34765625, "rewards/simpleverify_reward/std": 0.47669193148612976, "step": 263, "tools/generated_tokens": 2379.6484375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.552734375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.388671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1963.0, "completions/mean_length": 1423.111328125, "completions/mean_terminated_length": 1025.81787109375, "completions/min_length": 62.0, "completions/min_terminated_length": 62.0, "epoch": 0.1873669268985096, "frac_reward_zero_std": 0.28125, "grad_norm": 3.1716599464416504, "learning_rate": 4.7399999999999993e-07, "loss": 0.0612, "num_tokens": 626583998.0, "reward": 0.3515625, "reward_std": 0.2879050374031067, "rewards/simpleverify_reward/mean": 0.3515625, "rewards/simpleverify_reward/std": 0.4779251217842102, "step": 264, "tools/generated_tokens": 2490.123046875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.58984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.392578125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1924.0, "completions/mean_length": 1399.8828125, "completions/mean_terminated_length": 984.9420776367188, "completions/min_length": 33.0, "completions/min_terminated_length": 33.0, "epoch": 0.1880766501064585, "frac_reward_zero_std": 0.40625, "grad_norm": 4.451046466827393, "learning_rate": 4.7199999999999994e-07, "loss": 0.02, "num_tokens": 628664891.0, "reward": 0.271484375, "reward_std": 0.2479669600725174, "rewards/simpleverify_reward/mean": 0.271484375, "rewards/simpleverify_reward/std": 0.44516023993492126, "step": 265, "tools/generated_tokens": 2460.314453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.44921875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1943.0, "completions/mean_length": 1487.05859375, "completions/mean_terminated_length": 1035.9112548828125, "completions/min_length": 107.0, "completions/min_terminated_length": 107.0, "epoch": 0.18878637331440737, "frac_reward_zero_std": 0.34375, "grad_norm": 2.846372365951538, "learning_rate": 4.6999999999999995e-07, "loss": 0.0706, "num_tokens": 630878199.0, "reward": 0.35546875, "reward_std": 0.2538505494594574, "rewards/simpleverify_reward/mean": 0.35546875, "rewards/simpleverify_reward/std": 0.47912323474884033, "step": 266, "tools/generated_tokens": 2648.1953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.64453125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4609375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2001.0, "completions/mean_length": 1511.42578125, "completions/mean_terminated_length": 1053.3768310546875, "completions/min_length": 42.0, "completions/min_terminated_length": 42.0, "epoch": 0.18949609652235627, "frac_reward_zero_std": 0.15625, "grad_norm": 10.988624572753906, "learning_rate": 4.68e-07, "loss": 0.0788, "num_tokens": 633156300.0, "reward": 0.337890625, "reward_std": 0.3532556891441345, "rewards/simpleverify_reward/mean": 0.337890625, "rewards/simpleverify_reward/std": 0.4734536409378052, "step": 267, "tools/generated_tokens": 2727.955078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.41015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1983.0, "completions/mean_length": 1438.53515625, "completions/mean_terminated_length": 1014.735107421875, "completions/min_length": 202.0, "completions/min_terminated_length": 202.0, "epoch": 0.19020581973030518, "frac_reward_zero_std": 0.4375, "grad_norm": 7.3167829513549805, "learning_rate": 4.66e-07, "loss": 0.0459, "num_tokens": 635366463.0, "reward": 0.39453125, "reward_std": 0.2366800308227539, "rewards/simpleverify_reward/mean": 0.39453125, "rewards/simpleverify_reward/std": 0.4892277717590332, "step": 268, "tools/generated_tokens": 2631.181640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.650390625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.322265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1984.0, "completions/mean_length": 1394.138671875, "completions/mean_terminated_length": 1083.2247314453125, "completions/min_length": 41.0, "completions/min_terminated_length": 41.0, "epoch": 0.19091554293825408, "frac_reward_zero_std": 0.21875, "grad_norm": 2.7609825134277344, "learning_rate": 4.64e-07, "loss": 0.0425, "num_tokens": 637498491.0, "reward": 0.439453125, "reward_std": 0.3268650770187378, "rewards/simpleverify_reward/mean": 0.439453125, "rewards/simpleverify_reward/std": 0.49680593609809875, "step": 269, "tools/generated_tokens": 2512.3359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.615234375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.275390625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1986.0, "completions/mean_length": 1295.791015625, "completions/mean_terminated_length": 1010.47705078125, "completions/min_length": 122.0, "completions/min_terminated_length": 122.0, "epoch": 0.19162526614620298, "frac_reward_zero_std": 0.21875, "grad_norm": 16.956022262573242, "learning_rate": 4.62e-07, "loss": 0.0451, "num_tokens": 639459047.0, "reward": 0.5, "reward_std": 0.32369011640548706, "rewards/simpleverify_reward/mean": 0.5, "rewards/simpleverify_reward/std": 0.5004889965057373, "step": 270, "tools/generated_tokens": 2268.376953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.533203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.341796875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1961.0, "completions/mean_length": 1446.001953125, "completions/mean_terminated_length": 1134.6795654296875, "completions/min_length": 86.0, "completions/min_terminated_length": 86.0, "epoch": 0.1923349893541519, "frac_reward_zero_std": 0.25, "grad_norm": 7.749610424041748, "learning_rate": 4.6e-07, "loss": 0.0507, "num_tokens": 641564529.0, "reward": 0.4453125, "reward_std": 0.30862945318222046, "rewards/simpleverify_reward/mean": 0.4453125, "rewards/simpleverify_reward/std": 0.49748632311820984, "step": 271, "tools/generated_tokens": 2511.884765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.431640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1982.0, "completions/mean_length": 1544.162109375, "completions/mean_terminated_length": 1164.398681640625, "completions/min_length": 133.0, "completions/min_terminated_length": 133.0, "epoch": 0.1930447125621008, "frac_reward_zero_std": 0.40625, "grad_norm": 3.5215137004852295, "learning_rate": 4.58e-07, "loss": 0.0385, "num_tokens": 643883977.0, "reward": 0.37890625, "reward_std": 0.24058005213737488, "rewards/simpleverify_reward/mean": 0.37890625, "rewards/simpleverify_reward/std": 0.4855891764163971, "step": 272, "tools/generated_tokens": 2792.59765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.30859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1914.0, "completions/mean_length": 1364.98046875, "completions/mean_terminated_length": 1060.1300048828125, "completions/min_length": 147.0, "completions/min_terminated_length": 147.0, "epoch": 0.1937544357700497, "frac_reward_zero_std": 0.40625, "grad_norm": 3.8979361057281494, "learning_rate": 4.56e-07, "loss": 0.0316, "num_tokens": 645914384.0, "reward": 0.373046875, "reward_std": 0.2423821985721588, "rewards/simpleverify_reward/mean": 0.373046875, "rewards/simpleverify_reward/std": 0.48408737778663635, "step": 273, "tools/generated_tokens": 2373.6953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1919.0, "completions/mean_length": 1395.0, "completions/mean_terminated_length": 1040.9638671875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.19446415897799857, "frac_reward_zero_std": 0.3125, "grad_norm": 13.901947021484375, "learning_rate": 4.54e-07, "loss": 0.0514, "num_tokens": 648011758.0, "reward": 0.419921875, "reward_std": 0.2936236262321472, "rewards/simpleverify_reward/mean": 0.419921875, "rewards/simpleverify_reward/std": 0.4940285086631775, "step": 274, "tools/generated_tokens": 2469.78125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.357421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1978.0, "completions/mean_length": 1397.19921875, "completions/mean_terminated_length": 1035.20361328125, "completions/min_length": 43.0, "completions/min_terminated_length": 43.0, "epoch": 0.19517388218594747, "frac_reward_zero_std": 0.21875, "grad_norm": 12.363701820373535, "learning_rate": 4.5199999999999997e-07, "loss": 0.0553, "num_tokens": 650081499.0, "reward": 0.375, "reward_std": 0.2951202988624573, "rewards/simpleverify_reward/mean": 0.375, "rewards/simpleverify_reward/std": 0.4845963716506958, "step": 275, "tools/generated_tokens": 2444.4375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.580078125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.412109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1983.0, "completions/mean_length": 1489.115234375, "completions/mean_terminated_length": 1097.3388671875, "completions/min_length": 38.0, "completions/min_terminated_length": 38.0, "epoch": 0.19588360539389638, "frac_reward_zero_std": 0.25, "grad_norm": 4.517443656921387, "learning_rate": 4.5e-07, "loss": 0.0698, "num_tokens": 652384299.0, "reward": 0.34765625, "reward_std": 0.3019655644893646, "rewards/simpleverify_reward/mean": 0.34765625, "rewards/simpleverify_reward/std": 0.47669193148612976, "step": 276, "tools/generated_tokens": 2749.220703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.685546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.376953125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1998.0, "completions/mean_length": 1366.09765625, "completions/mean_terminated_length": 953.5955810546875, "completions/min_length": 76.0, "completions/min_terminated_length": 76.0, "epoch": 0.19659332860184528, "frac_reward_zero_std": 0.34375, "grad_norm": 6.681068420410156, "learning_rate": 4.48e-07, "loss": 0.0515, "num_tokens": 654427222.0, "reward": 0.365234375, "reward_std": 0.2513795495033264, "rewards/simpleverify_reward/mean": 0.365234375, "rewards/simpleverify_reward/std": 0.4819667339324951, "step": 277, "tools/generated_tokens": 2382.876953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1959.0, "completions/mean_length": 1452.873046875, "completions/mean_terminated_length": 1058.697998046875, "completions/min_length": 90.0, "completions/min_terminated_length": 90.0, "epoch": 0.19730305180979418, "frac_reward_zero_std": 0.28125, "grad_norm": 8.737835884094238, "learning_rate": 4.46e-07, "loss": 0.0133, "num_tokens": 656611295.0, "reward": 0.306640625, "reward_std": 0.26355257630348206, "rewards/simpleverify_reward/mean": 0.306640625, "rewards/simpleverify_reward/std": 0.4615498185157776, "step": 278, "tools/generated_tokens": 2600.9765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.630859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.41015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1962.0, "completions/mean_length": 1437.58203125, "completions/mean_terminated_length": 1013.31787109375, "completions/min_length": 130.0, "completions/min_terminated_length": 130.0, "epoch": 0.1980127750177431, "frac_reward_zero_std": 0.40625, "grad_norm": 6.756524562835693, "learning_rate": 4.44e-07, "loss": 0.0329, "num_tokens": 658749120.0, "reward": 0.412109375, "reward_std": 0.24670782685279846, "rewards/simpleverify_reward/mean": 0.412109375, "rewards/simpleverify_reward/std": 0.49269601702690125, "step": 279, "tools/generated_tokens": 2507.53125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.599609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.369140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1952.0, "completions/mean_length": 1444.236328125, "completions/mean_terminated_length": 1094.74609375, "completions/min_length": 113.0, "completions/min_terminated_length": 113.0, "epoch": 0.198722498225692, "frac_reward_zero_std": 0.21875, "grad_norm": 4.54815673828125, "learning_rate": 4.4199999999999996e-07, "loss": 0.0706, "num_tokens": 660951712.0, "reward": 0.435546875, "reward_std": 0.3011701703071594, "rewards/simpleverify_reward/mean": 0.435546875, "rewards/simpleverify_reward/std": 0.49631330370903015, "step": 280, "tools/generated_tokens": 2646.2421875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.65625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.439453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1970.0, "completions/mean_length": 1564.126953125, "completions/mean_terminated_length": 1184.783935546875, "completions/min_length": 88.0, "completions/min_terminated_length": 88.0, "epoch": 0.1994322214336409, "frac_reward_zero_std": 0.25, "grad_norm": 3.020799398422241, "learning_rate": 4.3999999999999997e-07, "loss": 0.0631, "num_tokens": 663352045.0, "reward": 0.3671875, "reward_std": 0.3239750862121582, "rewards/simpleverify_reward/mean": 0.3671875, "rewards/simpleverify_reward/std": 0.48250964283943176, "step": 281, "tools/generated_tokens": 2907.20703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.732421875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.44140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2008.0, "completions/mean_length": 1540.697265625, "completions/mean_terminated_length": 1140.4754638671875, "completions/min_length": 102.0, "completions/min_terminated_length": 102.0, "epoch": 0.20014194464158977, "frac_reward_zero_std": 0.34375, "grad_norm": 2.265376567840576, "learning_rate": 4.38e-07, "loss": 0.0623, "num_tokens": 665724876.0, "reward": 0.365234375, "reward_std": 0.2629948556423187, "rewards/simpleverify_reward/mean": 0.365234375, "rewards/simpleverify_reward/std": 0.4819667339324951, "step": 282, "tools/generated_tokens": 2863.291015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.197265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1966.0, "completions/mean_length": 1313.703125, "completions/mean_terminated_length": 1133.766357421875, "completions/min_length": 46.0, "completions/min_terminated_length": 46.0, "epoch": 0.20085166784953867, "frac_reward_zero_std": 0.21875, "grad_norm": 3.298328161239624, "learning_rate": 4.36e-07, "loss": 0.0585, "num_tokens": 667548823.0, "reward": 0.552734375, "reward_std": 0.33111709356307983, "rewards/simpleverify_reward/mean": 0.552734375, "rewards/simpleverify_reward/std": 0.4976975917816162, "step": 283, "tools/generated_tokens": 2149.97265625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.453125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4609375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1981.0, "completions/mean_length": 1528.041015625, "completions/mean_terminated_length": 1083.4384765625, "completions/min_length": 43.0, "completions/min_terminated_length": 43.0, "epoch": 0.20156139105748758, "frac_reward_zero_std": 0.28125, "grad_norm": 4.013789176940918, "learning_rate": 4.34e-07, "loss": 0.0784, "num_tokens": 669869271.0, "reward": 0.36328125, "reward_std": 0.3203025162220001, "rewards/simpleverify_reward/mean": 0.36328125, "rewards/simpleverify_reward/std": 0.4814152419567108, "step": 284, "tools/generated_tokens": 2802.142578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.697265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.365234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1987.0, "completions/mean_length": 1441.556640625, "completions/mean_terminated_length": 1093.9107666015625, "completions/min_length": 25.0, "completions/min_terminated_length": 25.0, "epoch": 0.20227111426543648, "frac_reward_zero_std": 0.21875, "grad_norm": 5.488464832305908, "learning_rate": 4.3199999999999995e-07, "loss": 0.0478, "num_tokens": 672024311.0, "reward": 0.3515625, "reward_std": 0.32739579677581787, "rewards/simpleverify_reward/mean": 0.3515625, "rewards/simpleverify_reward/std": 0.4779251217842102, "step": 285, "tools/generated_tokens": 2557.458984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.466796875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1988.0, "completions/mean_length": 1461.68359375, "completions/mean_terminated_length": 948.3883056640625, "completions/min_length": 89.0, "completions/min_terminated_length": 89.0, "epoch": 0.20298083747338538, "frac_reward_zero_std": 0.28125, "grad_norm": 4.03286600112915, "learning_rate": 4.2999999999999996e-07, "loss": 0.0604, "num_tokens": 674284054.0, "reward": 0.28125, "reward_std": 0.2734805941581726, "rewards/simpleverify_reward/mean": 0.28125, "rewards/simpleverify_reward/std": 0.45004892349243164, "step": 286, "tools/generated_tokens": 2704.861328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2008.0, "completions/mean_length": 1375.861328125, "completions/mean_terminated_length": 1081.32861328125, "completions/min_length": 33.0, "completions/min_terminated_length": 33.0, "epoch": 0.2036905606813343, "frac_reward_zero_std": 0.15625, "grad_norm": 4.266409873962402, "learning_rate": 4.2799999999999997e-07, "loss": 0.1049, "num_tokens": 676354786.0, "reward": 0.46484375, "reward_std": 0.3693467974662781, "rewards/simpleverify_reward/mean": 0.46484375, "rewards/simpleverify_reward/std": 0.49925029277801514, "step": 287, "tools/generated_tokens": 2454.27734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.587890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1986.0, "completions/mean_length": 1476.8984375, "completions/mean_terminated_length": 1074.8800048828125, "completions/min_length": 179.0, "completions/min_terminated_length": 179.0, "epoch": 0.2044002838892832, "frac_reward_zero_std": 0.375, "grad_norm": 3.800220012664795, "learning_rate": 4.26e-07, "loss": 0.0144, "num_tokens": 678506624.0, "reward": 0.3515625, "reward_std": 0.24834764003753662, "rewards/simpleverify_reward/mean": 0.3515625, "rewards/simpleverify_reward/std": 0.4779251217842102, "step": 288, "tools/generated_tokens": 2577.3203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.603515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.37109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1972.0, "completions/mean_length": 1435.880859375, "completions/mean_terminated_length": 1074.692626953125, "completions/min_length": 1.0, "completions/min_terminated_length": 1.0, "epoch": 0.20511000709723207, "frac_reward_zero_std": 0.25, "grad_norm": 2.352355718612671, "learning_rate": 4.24e-07, "loss": 0.066, "num_tokens": 680688587.0, "reward": 0.46484375, "reward_std": 0.2805333137512207, "rewards/simpleverify_reward/mean": 0.46484375, "rewards/simpleverify_reward/std": 0.49925029277801514, "step": 289, "tools/generated_tokens": 2588.919921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.630859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.41796875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1959.0, "completions/mean_length": 1410.224609375, "completions/mean_terminated_length": 952.224853515625, "completions/min_length": 75.0, "completions/min_terminated_length": 75.0, "epoch": 0.20581973030518097, "frac_reward_zero_std": 0.4375, "grad_norm": 2.8717739582061768, "learning_rate": 4.2199999999999994e-07, "loss": 0.071, "num_tokens": 682886665.0, "reward": 0.29296875, "reward_std": 0.21411387622356415, "rewards/simpleverify_reward/mean": 0.29296875, "rewards/simpleverify_reward/std": 0.455569326877594, "step": 290, "tools/generated_tokens": 2593.560546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.64453125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.361328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1958.0, "completions/mean_length": 1453.845703125, "completions/mean_terminated_length": 1120.3975830078125, "completions/min_length": 69.0, "completions/min_terminated_length": 69.0, "epoch": 0.20652945351312987, "frac_reward_zero_std": 0.28125, "grad_norm": 2.953495502471924, "learning_rate": 4.1999999999999995e-07, "loss": 0.0369, "num_tokens": 685068075.0, "reward": 0.38671875, "reward_std": 0.26667749881744385, "rewards/simpleverify_reward/mean": 0.38671875, "rewards/simpleverify_reward/std": 0.48747459053993225, "step": 291, "tools/generated_tokens": 2611.408203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.62890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.462890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1960.0, "completions/mean_length": 1569.830078125, "completions/mean_terminated_length": 1157.8145751953125, "completions/min_length": 24.0, "completions/min_terminated_length": 24.0, "epoch": 0.20723917672107878, "frac_reward_zero_std": 0.25, "grad_norm": 3.0336520671844482, "learning_rate": 4.1799999999999996e-07, "loss": 0.0521, "num_tokens": 687525686.0, "reward": 0.3125, "reward_std": 0.3385832607746124, "rewards/simpleverify_reward/mean": 0.3125, "rewards/simpleverify_reward/std": 0.4639657139778137, "step": 292, "tools/generated_tokens": 2983.1328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.76953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.2890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1959.0, "completions/mean_length": 1343.703125, "completions/mean_terminated_length": 1058.7884521484375, "completions/min_length": 91.0, "completions/min_terminated_length": 91.0, "epoch": 0.20794889992902768, "frac_reward_zero_std": 0.3125, "grad_norm": 2.597860336303711, "learning_rate": 4.1599999999999997e-07, "loss": 0.0938, "num_tokens": 689503953.0, "reward": 0.484375, "reward_std": 0.28495264053344727, "rewards/simpleverify_reward/mean": 0.484375, "rewards/simpleverify_reward/std": 0.5002445578575134, "step": 293, "tools/generated_tokens": 2314.787109375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.52734375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.2578125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1940.0, "completions/mean_length": 1297.158203125, "completions/mean_terminated_length": 1036.3394775390625, "completions/min_length": 30.0, "completions/min_terminated_length": 30.0, "epoch": 0.20865862313697658, "frac_reward_zero_std": 0.125, "grad_norm": 3.703139305114746, "learning_rate": 4.14e-07, "loss": 0.0461, "num_tokens": 691399741.0, "reward": 0.390625, "reward_std": 0.3479374647140503, "rewards/simpleverify_reward/mean": 0.390625, "rewards/simpleverify_reward/std": 0.48836761713027954, "step": 294, "tools/generated_tokens": 2196.41796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.49609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.36328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1929.0, "completions/mean_length": 1400.517578125, "completions/mean_terminated_length": 1032.4385986328125, "completions/min_length": 114.0, "completions/min_terminated_length": 114.0, "epoch": 0.2093683463449255, "frac_reward_zero_std": 0.34375, "grad_norm": 2.9751780033111572, "learning_rate": 4.12e-07, "loss": 0.0703, "num_tokens": 693523293.0, "reward": 0.455078125, "reward_std": 0.265807569026947, "rewards/simpleverify_reward/mean": 0.455078125, "rewards/simpleverify_reward/std": 0.4984649419784546, "step": 295, "tools/generated_tokens": 2513.5234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.353515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1952.0, "completions/mean_length": 1380.37890625, "completions/mean_terminated_length": 1015.8851928710938, "completions/min_length": 70.0, "completions/min_terminated_length": 70.0, "epoch": 0.2100780695528744, "frac_reward_zero_std": 0.28125, "grad_norm": 7.183781623840332, "learning_rate": 4.0999999999999994e-07, "loss": 0.0506, "num_tokens": 695598723.0, "reward": 0.369140625, "reward_std": 0.26569312810897827, "rewards/simpleverify_reward/mean": 0.369140625, "rewards/simpleverify_reward/std": 0.4830440282821655, "step": 296, "tools/generated_tokens": 2448.4375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.583984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1948.0, "completions/mean_length": 1459.955078125, "completions/mean_terminated_length": 1030.8411865234375, "completions/min_length": 88.0, "completions/min_terminated_length": 88.0, "epoch": 0.21078779276082327, "frac_reward_zero_std": 0.28125, "grad_norm": 8.137382507324219, "learning_rate": 4.0799999999999995e-07, "loss": 0.0659, "num_tokens": 697835622.0, "reward": 0.412109375, "reward_std": 0.29891422390937805, "rewards/simpleverify_reward/mean": 0.412109375, "rewards/simpleverify_reward/std": 0.49269601702690125, "step": 297, "tools/generated_tokens": 2671.666015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.384765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1970.0, "completions/mean_length": 1387.572265625, "completions/mean_terminated_length": 976.2095336914062, "completions/min_length": 23.0, "completions/min_terminated_length": 23.0, "epoch": 0.21149751596877217, "frac_reward_zero_std": 0.25, "grad_norm": 7.031959533691406, "learning_rate": 4.06e-07, "loss": 0.064, "num_tokens": 699904901.0, "reward": 0.34375, "reward_std": 0.3040506839752197, "rewards/simpleverify_reward/mean": 0.34375, "rewards/simpleverify_reward/std": 0.4754233956336975, "step": 298, "tools/generated_tokens": 2446.5234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.32421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2026.0, "completions/mean_length": 1391.17578125, "completions/mean_terminated_length": 1076.4970703125, "completions/min_length": 276.0, "completions/min_terminated_length": 276.0, "epoch": 0.21220723917672107, "frac_reward_zero_std": 0.28125, "grad_norm": 2.9769887924194336, "learning_rate": 4.04e-07, "loss": 0.0635, "num_tokens": 701924440.0, "reward": 0.498046875, "reward_std": 0.2843221426010132, "rewards/simpleverify_reward/mean": 0.498046875, "rewards/simpleverify_reward/std": 0.5004851818084717, "step": 299, "tools/generated_tokens": 2376.138671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.54296875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.337890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1980.0, "completions/mean_length": 1403.029296875, "completions/mean_terminated_length": 1073.885009765625, "completions/min_length": 62.0, "completions/min_terminated_length": 62.0, "epoch": 0.21291696238466998, "frac_reward_zero_std": 0.34375, "grad_norm": 3.0174896717071533, "learning_rate": 4.02e-07, "loss": 0.0415, "num_tokens": 704019821.0, "reward": 0.345703125, "reward_std": 0.27721232175827026, "rewards/simpleverify_reward/mean": 0.345703125, "rewards/simpleverify_reward/std": 0.4760620892047882, "step": 300, "tools/generated_tokens": 2494.591796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.595703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.48046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1957.0, "completions/mean_length": 1513.076171875, "completions/mean_terminated_length": 1018.3721923828125, "completions/min_length": 53.0, "completions/min_terminated_length": 53.0, "epoch": 0.21362668559261888, "frac_reward_zero_std": 0.4375, "grad_norm": 8.280428886413574, "learning_rate": 4e-07, "loss": 0.044, "num_tokens": 706314275.0, "reward": 0.306640625, "reward_std": 0.22427710890769958, "rewards/simpleverify_reward/mean": 0.306640625, "rewards/simpleverify_reward/std": 0.4615498185157776, "step": 301, "tools/generated_tokens": 2750.724609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.681640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.326171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1961.0, "completions/mean_length": 1348.189453125, "completions/mean_terminated_length": 1013.2435302734375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.21433640880056778, "frac_reward_zero_std": 0.28125, "grad_norm": 10.007421493530273, "learning_rate": 3.98e-07, "loss": 0.0466, "num_tokens": 708317470.0, "reward": 0.427734375, "reward_std": 0.28451114892959595, "rewards/simpleverify_reward/mean": 0.427734375, "rewards/simpleverify_reward/std": 0.4952339828014374, "step": 302, "tools/generated_tokens": 2334.787109375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.53515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.404296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1969.0, "completions/mean_length": 1453.8671875, "completions/mean_terminated_length": 1050.6361083984375, "completions/min_length": 75.0, "completions/min_terminated_length": 75.0, "epoch": 0.2150461320085167, "frac_reward_zero_std": 0.1875, "grad_norm": 13.425528526306152, "learning_rate": 3.96e-07, "loss": 0.0848, "num_tokens": 710544482.0, "reward": 0.4609375, "reward_std": 0.33255964517593384, "rewards/simpleverify_reward/mean": 0.4609375, "rewards/simpleverify_reward/std": 0.4989593029022217, "step": 303, "tools/generated_tokens": 2676.615234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.666015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.388671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2027.0, "completions/mean_length": 1437.87109375, "completions/mean_terminated_length": 1049.961669921875, "completions/min_length": 103.0, "completions/min_terminated_length": 103.0, "epoch": 0.2157558552164656, "frac_reward_zero_std": 0.21875, "grad_norm": 4.850423812866211, "learning_rate": 3.94e-07, "loss": 0.0382, "num_tokens": 712736558.0, "reward": 0.396484375, "reward_std": 0.3027869462966919, "rewards/simpleverify_reward/mean": 0.396484375, "rewards/simpleverify_reward/std": 0.4896455705165863, "step": 304, "tools/generated_tokens": 2614.029296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.638671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.36328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2025.0, "completions/mean_length": 1385.859375, "completions/mean_terminated_length": 1008.0736083984375, "completions/min_length": 47.0, "completions/min_terminated_length": 47.0, "epoch": 0.21646557842441447, "frac_reward_zero_std": 0.46875, "grad_norm": 4.820125102996826, "learning_rate": 3.92e-07, "loss": 0.0396, "num_tokens": 714926967.0, "reward": 0.373046875, "reward_std": 0.22742411494255066, "rewards/simpleverify_reward/mean": 0.373046875, "rewards/simpleverify_reward/std": 0.48408737778663635, "step": 305, "tools/generated_tokens": 2585.04296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.341796875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1973.0, "completions/mean_length": 1363.982421875, "completions/mean_terminated_length": 1010.6320190429688, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.21717530163236337, "frac_reward_zero_std": 0.1875, "grad_norm": 5.822154998779297, "learning_rate": 3.8999999999999997e-07, "loss": 0.0426, "num_tokens": 716989320.0, "reward": 0.46875, "reward_std": 0.35226404666900635, "rewards/simpleverify_reward/mean": 0.46875, "rewards/simpleverify_reward/std": 0.4995105266571045, "step": 306, "tools/generated_tokens": 2413.0078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.373046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1968.0, "completions/mean_length": 1414.83203125, "completions/mean_terminated_length": 1038.087158203125, "completions/min_length": 42.0, "completions/min_terminated_length": 42.0, "epoch": 0.21788502484031227, "frac_reward_zero_std": 0.40625, "grad_norm": 2.306669235229492, "learning_rate": 3.88e-07, "loss": 0.0316, "num_tokens": 719065796.0, "reward": 0.3828125, "reward_std": 0.25882071256637573, "rewards/simpleverify_reward/mean": 0.3828125, "rewards/simpleverify_reward/std": 0.486548513174057, "step": 307, "tools/generated_tokens": 2424.1640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.564453125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.283203125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1991.0, "completions/mean_length": 1256.90625, "completions/mean_terminated_length": 945.52587890625, "completions/min_length": 92.0, "completions/min_terminated_length": 92.0, "epoch": 0.21859474804826118, "frac_reward_zero_std": 0.28125, "grad_norm": 3.96085524559021, "learning_rate": 3.86e-07, "loss": 0.0807, "num_tokens": 720940572.0, "reward": 0.373046875, "reward_std": 0.31049197912216187, "rewards/simpleverify_reward/mean": 0.373046875, "rewards/simpleverify_reward/std": 0.48408737778663635, "step": 308, "tools/generated_tokens": 2162.6171875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.49609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.322265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1946.0, "completions/mean_length": 1423.560546875, "completions/mean_terminated_length": 1127.305419921875, "completions/min_length": 77.0, "completions/min_terminated_length": 77.0, "epoch": 0.21930447125621008, "frac_reward_zero_std": 0.1875, "grad_norm": 3.7559847831726074, "learning_rate": 3.84e-07, "loss": 0.0502, "num_tokens": 723029561.0, "reward": 0.4609375, "reward_std": 0.35038772225379944, "rewards/simpleverify_reward/mean": 0.4609375, "rewards/simpleverify_reward/std": 0.4989593029022217, "step": 309, "tools/generated_tokens": 2482.0546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.576171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.41015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1969.0, "completions/mean_length": 1473.953125, "completions/mean_terminated_length": 1075.41064453125, "completions/min_length": 101.0, "completions/min_terminated_length": 101.0, "epoch": 0.22001419446415899, "frac_reward_zero_std": 0.21875, "grad_norm": 3.7668988704681396, "learning_rate": 3.82e-07, "loss": 0.06, "num_tokens": 725204948.0, "reward": 0.41015625, "reward_std": 0.2874017655849457, "rewards/simpleverify_reward/mean": 0.41015625, "rewards/simpleverify_reward/std": 0.49234291911125183, "step": 310, "tools/generated_tokens": 2598.904296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.61328125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.326171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1951.0, "completions/mean_length": 1324.728515625, "completions/mean_terminated_length": 976.2000122070312, "completions/min_length": 125.0, "completions/min_terminated_length": 125.0, "epoch": 0.2207239176721079, "frac_reward_zero_std": 0.21875, "grad_norm": 5.326725482940674, "learning_rate": 3.7999999999999996e-07, "loss": 0.0565, "num_tokens": 727188384.0, "reward": 0.4140625, "reward_std": 0.3073521852493286, "rewards/simpleverify_reward/mean": 0.4140625, "rewards/simpleverify_reward/std": 0.49304109811782837, "step": 311, "tools/generated_tokens": 2308.513671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.53515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.443359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1997.0, "completions/mean_length": 1477.759765625, "completions/mean_terminated_length": 1023.5684204101562, "completions/min_length": 74.0, "completions/min_terminated_length": 74.0, "epoch": 0.22143364088005676, "frac_reward_zero_std": 0.375, "grad_norm": 2.3207380771636963, "learning_rate": 3.7799999999999997e-07, "loss": 0.0671, "num_tokens": 729440821.0, "reward": 0.3984375, "reward_std": 0.25145286321640015, "rewards/simpleverify_reward/mean": 0.3984375, "rewards/simpleverify_reward/std": 0.4900552034378052, "step": 312, "tools/generated_tokens": 2683.578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.2421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1969.0, "completions/mean_length": 1280.046875, "completions/mean_terminated_length": 1036.0721435546875, "completions/min_length": 92.0, "completions/min_terminated_length": 92.0, "epoch": 0.22214336408800567, "frac_reward_zero_std": 0.3125, "grad_norm": 5.503763675689697, "learning_rate": 3.76e-07, "loss": 0.0203, "num_tokens": 731246088.0, "reward": 0.453125, "reward_std": 0.2584003210067749, "rewards/simpleverify_reward/mean": 0.453125, "rewards/simpleverify_reward/std": 0.4982847273349762, "step": 313, "tools/generated_tokens": 2082.068359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.44140625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.396484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1943.0, "completions/mean_length": 1390.265625, "completions/mean_terminated_length": 958.33984375, "completions/min_length": 49.0, "completions/min_terminated_length": 49.0, "epoch": 0.22285308729595457, "frac_reward_zero_std": 0.21875, "grad_norm": 5.461803436279297, "learning_rate": 3.74e-07, "loss": 0.0406, "num_tokens": 733337579.0, "reward": 0.326171875, "reward_std": 0.2864355444908142, "rewards/simpleverify_reward/mean": 0.326171875, "rewards/simpleverify_reward/std": 0.4692695140838623, "step": 314, "tools/generated_tokens": 2451.9296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.587890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.34765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2011.0, "completions/mean_length": 1371.642578125, "completions/mean_terminated_length": 1011.1886596679688, "completions/min_length": 1.0, "completions/min_terminated_length": 1.0, "epoch": 0.22356281050390348, "frac_reward_zero_std": 0.25, "grad_norm": 3.585106611251831, "learning_rate": 3.72e-07, "loss": 0.0683, "num_tokens": 735398120.0, "reward": 0.40234375, "reward_std": 0.30691760778427124, "rewards/simpleverify_reward/mean": 0.40234375, "rewards/simpleverify_reward/std": 0.4908501207828522, "step": 315, "tools/generated_tokens": 2429.1875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.576171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1969.0, "completions/mean_length": 1351.640625, "completions/mean_terminated_length": 1047.6068115234375, "completions/min_length": 164.0, "completions/min_terminated_length": 164.0, "epoch": 0.22427253371185238, "frac_reward_zero_std": 0.1875, "grad_norm": 4.802496910095215, "learning_rate": 3.7e-07, "loss": 0.0651, "num_tokens": 737507832.0, "reward": 0.509765625, "reward_std": 0.3189505934715271, "rewards/simpleverify_reward/mean": 0.509765625, "rewards/simpleverify_reward/std": 0.5003935098648071, "step": 316, "tools/generated_tokens": 2466.525390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.60546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.279296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1964.0, "completions/mean_length": 1375.798828125, "completions/mean_terminated_length": 1115.8184814453125, "completions/min_length": 37.0, "completions/min_terminated_length": 37.0, "epoch": 0.22498225691980128, "frac_reward_zero_std": 0.15625, "grad_norm": 2.42252254486084, "learning_rate": 3.6799999999999996e-07, "loss": 0.0673, "num_tokens": 739521333.0, "reward": 0.509765625, "reward_std": 0.3291349411010742, "rewards/simpleverify_reward/mean": 0.509765625, "rewards/simpleverify_reward/std": 0.5003935098648071, "step": 317, "tools/generated_tokens": 2371.689453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.541015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1986.0, "completions/mean_length": 1497.958984375, "completions/mean_terminated_length": 1042.210693359375, "completions/min_length": 82.0, "completions/min_terminated_length": 82.0, "epoch": 0.22569198012775019, "frac_reward_zero_std": 0.21875, "grad_norm": 8.27332592010498, "learning_rate": 3.6599999999999997e-07, "loss": 0.0609, "num_tokens": 741821515.0, "reward": 0.357421875, "reward_std": 0.3280338644981384, "rewards/simpleverify_reward/mean": 0.357421875, "rewards/simpleverify_reward/std": 0.4797092080116272, "step": 318, "tools/generated_tokens": 2760.984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.69140625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.333984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1941.0, "completions/mean_length": 1410.291015625, "completions/mean_terminated_length": 1090.739013671875, "completions/min_length": 161.0, "completions/min_terminated_length": 161.0, "epoch": 0.2264017033356991, "frac_reward_zero_std": 0.21875, "grad_norm": 8.752574920654297, "learning_rate": 3.64e-07, "loss": 0.0897, "num_tokens": 743992409.0, "reward": 0.443359375, "reward_std": 0.32874634861946106, "rewards/simpleverify_reward/mean": 0.443359375, "rewards/simpleverify_reward/std": 0.49726733565330505, "step": 319, "tools/generated_tokens": 2576.771484375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6328125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.310546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1922.0, "completions/mean_length": 1277.162109375, "completions/mean_terminated_length": 930.8045043945312, "completions/min_length": 56.0, "completions/min_terminated_length": 56.0, "epoch": 0.22711142654364797, "frac_reward_zero_std": 0.375, "grad_norm": 2.5520641803741455, "learning_rate": 3.62e-07, "loss": 0.0581, "num_tokens": 745933354.0, "reward": 0.462890625, "reward_std": 0.2369387149810791, "rewards/simpleverify_reward/mean": 0.462890625, "rewards/simpleverify_reward/std": 0.4991086423397064, "step": 320, "tools/generated_tokens": 2252.91796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.533203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.26953125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2029.0, "completions/mean_length": 1324.234375, "completions/mean_terminated_length": 1057.176513671875, "completions/min_length": 46.0, "completions/min_terminated_length": 46.0, "epoch": 0.22782114975159687, "frac_reward_zero_std": 0.1875, "grad_norm": 11.809249877929688, "learning_rate": 3.6e-07, "loss": 0.0795, "num_tokens": 747896580.0, "reward": 0.40625, "reward_std": 0.3289005160331726, "rewards/simpleverify_reward/mean": 0.40625, "rewards/simpleverify_reward/std": 0.49161264300346375, "step": 321, "tools/generated_tokens": 2314.689453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.53515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.45703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1977.0, "completions/mean_length": 1531.232421875, "completions/mean_terminated_length": 1097.8165283203125, "completions/min_length": 39.0, "completions/min_terminated_length": 39.0, "epoch": 0.22853087295954577, "frac_reward_zero_std": 0.28125, "grad_norm": 11.850707054138184, "learning_rate": 3.5799999999999995e-07, "loss": 0.0369, "num_tokens": 750276043.0, "reward": 0.353515625, "reward_std": 0.3004316985607147, "rewards/simpleverify_reward/mean": 0.353515625, "rewards/simpleverify_reward/std": 0.47852855920791626, "step": 322, "tools/generated_tokens": 2866.626953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.728515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1999.0, "completions/mean_length": 1409.03125, "completions/mean_terminated_length": 1013.8164672851562, "completions/min_length": 136.0, "completions/min_terminated_length": 136.0, "epoch": 0.22924059616749468, "frac_reward_zero_std": 0.34375, "grad_norm": 7.15058708190918, "learning_rate": 3.5599999999999996e-07, "loss": 0.086, "num_tokens": 752430996.0, "reward": 0.318359375, "reward_std": 0.25130030512809753, "rewards/simpleverify_reward/mean": 0.318359375, "rewards/simpleverify_reward/std": 0.46629536151885986, "step": 323, "tools/generated_tokens": 2539.861328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.62109375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.365234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1956.0, "completions/mean_length": 1414.97265625, "completions/mean_terminated_length": 1050.7384033203125, "completions/min_length": 121.0, "completions/min_terminated_length": 121.0, "epoch": 0.22995031937544358, "frac_reward_zero_std": 0.21875, "grad_norm": 3.4098055362701416, "learning_rate": 3.5399999999999997e-07, "loss": 0.0752, "num_tokens": 754637475.0, "reward": 0.3984375, "reward_std": 0.3338572382926941, "rewards/simpleverify_reward/mean": 0.3984375, "rewards/simpleverify_reward/std": 0.4900552034378052, "step": 324, "tools/generated_tokens": 2615.953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.654296875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1972.0, "completions/mean_length": 1442.67578125, "completions/mean_terminated_length": 1103.1036376953125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.23066004258339248, "frac_reward_zero_std": 0.3125, "grad_norm": 6.135842323303223, "learning_rate": 3.52e-07, "loss": 0.0624, "num_tokens": 756843036.0, "reward": 0.333984375, "reward_std": 0.2735273838043213, "rewards/simpleverify_reward/mean": 0.333984375, "rewards/simpleverify_reward/std": 0.47209542989730835, "step": 325, "tools/generated_tokens": 2645.3359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.654296875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.46484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1933.0, "completions/mean_length": 1568.544921875, "completions/mean_terminated_length": 1152.083984375, "completions/min_length": 145.0, "completions/min_terminated_length": 145.0, "epoch": 0.2313697657913414, "frac_reward_zero_std": 0.25, "grad_norm": 9.425222396850586, "learning_rate": 3.5e-07, "loss": 0.0814, "num_tokens": 759191913.0, "reward": 0.291015625, "reward_std": 0.33528950810432434, "rewards/simpleverify_reward/mean": 0.291015625, "rewards/simpleverify_reward/std": 0.45467492938041687, "step": 326, "tools/generated_tokens": 2833.203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1897.0, "completions/mean_length": 1306.033203125, "completions/mean_terminated_length": 930.87060546875, "completions/min_length": 115.0, "completions/min_terminated_length": 115.0, "epoch": 0.2320794889992903, "frac_reward_zero_std": 0.28125, "grad_norm": 1.9694534540176392, "learning_rate": 3.4799999999999994e-07, "loss": 0.0584, "num_tokens": 761121147.0, "reward": 0.46484375, "reward_std": 0.3004428446292877, "rewards/simpleverify_reward/mean": 0.46484375, "rewards/simpleverify_reward/std": 0.49925029277801514, "step": 327, "tools/generated_tokens": 2238.9296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.51171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.310546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1994.0, "completions/mean_length": 1316.79296875, "completions/mean_terminated_length": 988.6883544921875, "completions/min_length": 83.0, "completions/min_terminated_length": 83.0, "epoch": 0.23278921220723917, "frac_reward_zero_std": 0.3125, "grad_norm": 2.998008966445923, "learning_rate": 3.4599999999999995e-07, "loss": 0.0372, "num_tokens": 763105367.0, "reward": 0.388671875, "reward_std": 0.27516451478004456, "rewards/simpleverify_reward/mean": 0.388671875, "rewards/simpleverify_reward/std": 0.4879252314567566, "step": 328, "tools/generated_tokens": 2305.064453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.541015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.380859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1965.0, "completions/mean_length": 1391.564453125, "completions/mean_terminated_length": 987.763427734375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.23349893541518807, "frac_reward_zero_std": 0.3125, "grad_norm": 4.8110880851745605, "learning_rate": 3.4399999999999996e-07, "loss": 0.0402, "num_tokens": 765189622.0, "reward": 0.369140625, "reward_std": 0.28113144636154175, "rewards/simpleverify_reward/mean": 0.369140625, "rewards/simpleverify_reward/std": 0.4830440282821655, "step": 329, "tools/generated_tokens": 2461.98828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.58984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1921.0, "completions/mean_length": 1527.763671875, "completions/mean_terminated_length": 1099.210693359375, "completions/min_length": 55.0, "completions/min_terminated_length": 55.0, "epoch": 0.23420865862313697, "frac_reward_zero_std": 0.25, "grad_norm": 2.0677883625030518, "learning_rate": 3.42e-07, "loss": 0.0696, "num_tokens": 767511873.0, "reward": 0.27734375, "reward_std": 0.2874564528465271, "rewards/simpleverify_reward/mean": 0.27734375, "rewards/simpleverify_reward/std": 0.4481254518032074, "step": 330, "tools/generated_tokens": 2805.64453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.69921875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.447265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1965.0, "completions/mean_length": 1506.23828125, "completions/mean_terminated_length": 1068.8056640625, "completions/min_length": 77.0, "completions/min_terminated_length": 77.0, "epoch": 0.23491838183108588, "frac_reward_zero_std": 0.375, "grad_norm": 6.645218849182129, "learning_rate": 3.4000000000000003e-07, "loss": 0.0337, "num_tokens": 769787992.0, "reward": 0.298828125, "reward_std": 0.2657744586467743, "rewards/simpleverify_reward/mean": 0.298828125, "rewards/simpleverify_reward/std": 0.45819199085235596, "step": 331, "tools/generated_tokens": 2711.33203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.314453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1958.0, "completions/mean_length": 1360.44921875, "completions/mean_terminated_length": 1045.076904296875, "completions/min_length": 39.0, "completions/min_terminated_length": 39.0, "epoch": 0.23562810503903478, "frac_reward_zero_std": 0.3125, "grad_norm": 3.6829135417938232, "learning_rate": 3.38e-07, "loss": 0.0213, "num_tokens": 771850576.0, "reward": 0.41015625, "reward_std": 0.2691202163696289, "rewards/simpleverify_reward/mean": 0.41015625, "rewards/simpleverify_reward/std": 0.49234291911125183, "step": 332, "tools/generated_tokens": 2408.123046875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.56640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2001.0, "completions/mean_length": 1345.81640625, "completions/mean_terminated_length": 1002.8895263671875, "completions/min_length": 51.0, "completions/min_terminated_length": 51.0, "epoch": 0.23633782824698368, "frac_reward_zero_std": 0.1875, "grad_norm": 2.9292025566101074, "learning_rate": 3.36e-07, "loss": 0.0458, "num_tokens": 773891048.0, "reward": 0.484375, "reward_std": 0.3180251121520996, "rewards/simpleverify_reward/mean": 0.484375, "rewards/simpleverify_reward/std": 0.5002445578575134, "step": 333, "tools/generated_tokens": 2369.556640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1953.0, "completions/mean_length": 1426.8515625, "completions/mean_terminated_length": 1054.1875, "completions/min_length": 158.0, "completions/min_terminated_length": 158.0, "epoch": 0.2370475514549326, "frac_reward_zero_std": 0.40625, "grad_norm": 1.623494267463684, "learning_rate": 3.34e-07, "loss": 0.0834, "num_tokens": 775990620.0, "reward": 0.359375, "reward_std": 0.24973651766777039, "rewards/simpleverify_reward/mean": 0.359375, "rewards/simpleverify_reward/std": 0.48028653860092163, "step": 334, "tools/generated_tokens": 2490.87890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.58984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.298828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2010.0, "completions/mean_length": 1360.814453125, "completions/mean_terminated_length": 1067.947021484375, "completions/min_length": 29.0, "completions/min_terminated_length": 29.0, "epoch": 0.23775727466288146, "frac_reward_zero_std": 0.25, "grad_norm": 2.3581292629241943, "learning_rate": 3.32e-07, "loss": 0.0471, "num_tokens": 778072637.0, "reward": 0.47265625, "reward_std": 0.2866290807723999, "rewards/simpleverify_reward/mean": 0.47265625, "rewards/simpleverify_reward/std": 0.49974003434181213, "step": 335, "tools/generated_tokens": 2453.3046875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.58984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.33984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1965.0, "completions/mean_length": 1428.568359375, "completions/mean_terminated_length": 1109.6893310546875, "completions/min_length": 41.0, "completions/min_terminated_length": 41.0, "epoch": 0.23846699787083037, "frac_reward_zero_std": 0.34375, "grad_norm": 4.168668746948242, "learning_rate": 3.3e-07, "loss": 0.0534, "num_tokens": 780152304.0, "reward": 0.40234375, "reward_std": 0.2882212996482849, "rewards/simpleverify_reward/mean": 0.40234375, "rewards/simpleverify_reward/std": 0.4908501207828522, "step": 336, "tools/generated_tokens": 2441.611328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.36328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1996.0, "completions/mean_length": 1412.20703125, "completions/mean_terminated_length": 1050.420166015625, "completions/min_length": 26.0, "completions/min_terminated_length": 26.0, "epoch": 0.23917672107877927, "frac_reward_zero_std": 0.3125, "grad_norm": 2.8757333755493164, "learning_rate": 3.28e-07, "loss": 0.0417, "num_tokens": 782453284.0, "reward": 0.3046875, "reward_std": 0.2878001928329468, "rewards/simpleverify_reward/mean": 0.3046875, "rewards/simpleverify_reward/std": 0.4607250988483429, "step": 337, "tools/generated_tokens": 2739.58984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.724609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.37890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1945.0, "completions/mean_length": 1437.3125, "completions/mean_terminated_length": 1065.7137451171875, "completions/min_length": 113.0, "completions/min_terminated_length": 113.0, "epoch": 0.23988644428672817, "frac_reward_zero_std": 0.3125, "grad_norm": 2.333538770675659, "learning_rate": 3.26e-07, "loss": 0.0353, "num_tokens": 784669797.0, "reward": 0.328125, "reward_std": 0.28373032808303833, "rewards/simpleverify_reward/mean": 0.328125, "rewards/simpleverify_reward/std": 0.4699897766113281, "step": 338, "tools/generated_tokens": 2635.07421875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.658203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.388671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1940.0, "completions/mean_length": 1419.6171875, "completions/mean_terminated_length": 1020.1022338867188, "completions/min_length": 163.0, "completions/min_terminated_length": 163.0, "epoch": 0.24059616749467708, "frac_reward_zero_std": 0.34375, "grad_norm": 3.102073907852173, "learning_rate": 3.24e-07, "loss": 0.0656, "num_tokens": 786858775.0, "reward": 0.359375, "reward_std": 0.275622695684433, "rewards/simpleverify_reward/mean": 0.359375, "rewards/simpleverify_reward/std": 0.48028653860092163, "step": 339, "tools/generated_tokens": 2605.48828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.654296875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1910.0, "completions/mean_length": 1396.728515625, "completions/mean_terminated_length": 1005.9656372070312, "completions/min_length": 112.0, "completions/min_terminated_length": 112.0, "epoch": 0.24130589070262598, "frac_reward_zero_std": 0.3125, "grad_norm": 1.3386632204055786, "learning_rate": 3.22e-07, "loss": 0.0371, "num_tokens": 788980922.0, "reward": 0.37109375, "reward_std": 0.28521260619163513, "rewards/simpleverify_reward/mean": 0.37109375, "rewards/simpleverify_reward/std": 0.4835699498653412, "step": 340, "tools/generated_tokens": 2502.181640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.365234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1975.0, "completions/mean_length": 1410.9375, "completions/mean_terminated_length": 1044.8245849609375, "completions/min_length": 110.0, "completions/min_terminated_length": 110.0, "epoch": 0.24201561391057488, "frac_reward_zero_std": 0.21875, "grad_norm": 4.672500133514404, "learning_rate": 3.2e-07, "loss": 0.0669, "num_tokens": 791134330.0, "reward": 0.34375, "reward_std": 0.3038235306739807, "rewards/simpleverify_reward/mean": 0.34375, "rewards/simpleverify_reward/std": 0.4754233956336975, "step": 341, "tools/generated_tokens": 2554.91015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.630859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.306640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2047.0, "completions/mean_length": 1358.8984375, "completions/mean_terminated_length": 1057.6055908203125, "completions/min_length": 116.0, "completions/min_terminated_length": 116.0, "epoch": 0.2427253371185238, "frac_reward_zero_std": 0.28125, "grad_norm": 4.280205249786377, "learning_rate": 3.18e-07, "loss": 0.0471, "num_tokens": 793117803.0, "reward": 0.435546875, "reward_std": 0.32077133655548096, "rewards/simpleverify_reward/mean": 0.435546875, "rewards/simpleverify_reward/std": 0.49631330370903015, "step": 342, "tools/generated_tokens": 2336.328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.533203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.439453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1961.0, "completions/mean_length": 1492.66015625, "completions/mean_terminated_length": 1060.9755859375, "completions/min_length": 116.0, "completions/min_terminated_length": 116.0, "epoch": 0.24343506032647266, "frac_reward_zero_std": 0.375, "grad_norm": 2.1949691772460938, "learning_rate": 3.1599999999999997e-07, "loss": 0.0295, "num_tokens": 795422879.0, "reward": 0.380859375, "reward_std": 0.27711838483810425, "rewards/simpleverify_reward/mean": 0.380859375, "rewards/simpleverify_reward/std": 0.48607301712036133, "step": 343, "tools/generated_tokens": 2756.390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.69921875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2004.0, "completions/mean_length": 1359.830078125, "completions/mean_terminated_length": 1011.6970825195312, "completions/min_length": 104.0, "completions/min_terminated_length": 104.0, "epoch": 0.24414478353442157, "frac_reward_zero_std": 0.25, "grad_norm": 4.722331523895264, "learning_rate": 3.14e-07, "loss": 0.0541, "num_tokens": 797533830.0, "reward": 0.439453125, "reward_std": 0.2888985872268677, "rewards/simpleverify_reward/mean": 0.439453125, "rewards/simpleverify_reward/std": 0.49680593609809875, "step": 344, "tools/generated_tokens": 2467.337890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.60546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.40625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1946.0, "completions/mean_length": 1423.03125, "completions/mean_terminated_length": 996.7368774414062, "completions/min_length": 103.0, "completions/min_terminated_length": 103.0, "epoch": 0.24485450674237047, "frac_reward_zero_std": 0.3125, "grad_norm": 12.235193252563477, "learning_rate": 3.12e-07, "loss": 0.0211, "num_tokens": 799712475.0, "reward": 0.34375, "reward_std": 0.27683311700820923, "rewards/simpleverify_reward/mean": 0.34375, "rewards/simpleverify_reward/std": 0.4754233956336975, "step": 345, "tools/generated_tokens": 2574.064453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.634765625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.33984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2000.0, "completions/mean_length": 1411.939453125, "completions/mean_terminated_length": 1086.872802734375, "completions/min_length": 38.0, "completions/min_terminated_length": 38.0, "epoch": 0.24556422995031937, "frac_reward_zero_std": 0.21875, "grad_norm": 3.868318796157837, "learning_rate": 3.1e-07, "loss": 0.0746, "num_tokens": 801908379.0, "reward": 0.3671875, "reward_std": 0.3346020579338074, "rewards/simpleverify_reward/mean": 0.3671875, "rewards/simpleverify_reward/std": 0.48250964283943176, "step": 346, "tools/generated_tokens": 2617.05859375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.650390625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2005.0, "completions/mean_length": 1358.447265625, "completions/mean_terminated_length": 1045.0142822265625, "completions/min_length": 123.0, "completions/min_terminated_length": 123.0, "epoch": 0.24627395315826828, "frac_reward_zero_std": 0.375, "grad_norm": 5.526366233825684, "learning_rate": 3.08e-07, "loss": 0.0557, "num_tokens": 803900849.0, "reward": 0.359375, "reward_std": 0.2672569155693054, "rewards/simpleverify_reward/mean": 0.359375, "rewards/simpleverify_reward/std": 0.48028653860092163, "step": 347, "tools/generated_tokens": 2348.95703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5390625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.40234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2015.0, "completions/mean_length": 1457.701171875, "completions/mean_terminated_length": 1060.3104248046875, "completions/min_length": 86.0, "completions/min_terminated_length": 86.0, "epoch": 0.24698367636621718, "frac_reward_zero_std": 0.3125, "grad_norm": 5.064498424530029, "learning_rate": 3.0599999999999996e-07, "loss": 0.0556, "num_tokens": 806223211.0, "reward": 0.3203125, "reward_std": 0.27993449568748474, "rewards/simpleverify_reward/mean": 0.3203125, "rewards/simpleverify_reward/std": 0.4670529365539551, "step": 348, "tools/generated_tokens": 2760.857421875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.71484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.478515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1957.0, "completions/mean_length": 1515.03515625, "completions/mean_terminated_length": 1025.9849853515625, "completions/min_length": 30.0, "completions/min_terminated_length": 30.0, "epoch": 0.24769339957416608, "frac_reward_zero_std": 0.375, "grad_norm": 1.7405883073806763, "learning_rate": 3.0399999999999997e-07, "loss": 0.0439, "num_tokens": 808542334.0, "reward": 0.3203125, "reward_std": 0.24359987676143646, "rewards/simpleverify_reward/mean": 0.3203125, "rewards/simpleverify_reward/std": 0.4670529365539551, "step": 349, "tools/generated_tokens": 2778.8984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.693359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.40234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1973.0, "completions/mean_length": 1474.5390625, "completions/mean_terminated_length": 1088.483642578125, "completions/min_length": 46.0, "completions/min_terminated_length": 46.0, "epoch": 0.248403122782115, "frac_reward_zero_std": 0.25, "grad_norm": 4.030844211578369, "learning_rate": 3.02e-07, "loss": 0.0684, "num_tokens": 810869557.0, "reward": 0.33203125, "reward_std": 0.3079848885536194, "rewards/simpleverify_reward/mean": 0.33203125, "rewards/simpleverify_reward/std": 0.47140273451805115, "step": 350, "tools/generated_tokens": 2786.86328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.712890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.2734375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1984.0, "completions/mean_length": 1277.111328125, "completions/mean_terminated_length": 986.991943359375, "completions/min_length": 81.0, "completions/min_terminated_length": 81.0, "epoch": 0.24911284599006386, "frac_reward_zero_std": 0.3125, "grad_norm": 3.2583823204040527, "learning_rate": 3e-07, "loss": 0.0455, "num_tokens": 812694793.0, "reward": 0.525390625, "reward_std": 0.24913008511066437, "rewards/simpleverify_reward/mean": 0.525390625, "rewards/simpleverify_reward/std": 0.4998432695865631, "step": 351, "tools/generated_tokens": 2107.583984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.455078125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.33984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1966.0, "completions/mean_length": 1329.412109375, "completions/mean_terminated_length": 959.4881591796875, "completions/min_length": 181.0, "completions/min_terminated_length": 181.0, "epoch": 0.24982256919801277, "frac_reward_zero_std": 0.28125, "grad_norm": 4.087608814239502, "learning_rate": 2.98e-07, "loss": 0.102, "num_tokens": 814687475.0, "reward": 0.41015625, "reward_std": 0.2839886546134949, "rewards/simpleverify_reward/mean": 0.41015625, "rewards/simpleverify_reward/std": 0.49234291911125183, "step": 352, "tools/generated_tokens": 2337.587890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.548828125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.44140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1957.0, "completions/mean_length": 1478.3828125, "completions/mean_terminated_length": 1028.2657470703125, "completions/min_length": 64.0, "completions/min_terminated_length": 64.0, "epoch": 0.2505322924059617, "frac_reward_zero_std": 0.3125, "grad_norm": 4.775875091552734, "learning_rate": 2.9599999999999995e-07, "loss": 0.0435, "num_tokens": 817011917.0, "reward": 0.32421875, "reward_std": 0.27779942750930786, "rewards/simpleverify_reward/mean": 0.32421875, "rewards/simpleverify_reward/std": 0.4685399830341339, "step": 353, "tools/generated_tokens": 2800.23828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.71875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.28515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1995.0, "completions/mean_length": 1320.763671875, "completions/mean_terminated_length": 1030.6639404296875, "completions/min_length": 2.0, "completions/min_terminated_length": 2.0, "epoch": 0.2512420156139106, "frac_reward_zero_std": 0.3125, "grad_norm": 5.358983993530273, "learning_rate": 2.9399999999999996e-07, "loss": 0.0562, "num_tokens": 818935007.0, "reward": 0.478515625, "reward_std": 0.283348023891449, "rewards/simpleverify_reward/mean": 0.478515625, "rewards/simpleverify_reward/std": 0.5000267624855042, "step": 354, "tools/generated_tokens": 2251.716796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.50390625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.404296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1974.0, "completions/mean_length": 1458.68359375, "completions/mean_terminated_length": 1058.7213134765625, "completions/min_length": 46.0, "completions/min_terminated_length": 46.0, "epoch": 0.25195173882185945, "frac_reward_zero_std": 0.1875, "grad_norm": 3.502298593521118, "learning_rate": 2.9199999999999997e-07, "loss": 0.0685, "num_tokens": 821176131.0, "reward": 0.412109375, "reward_std": 0.3489822447299957, "rewards/simpleverify_reward/mean": 0.412109375, "rewards/simpleverify_reward/std": 0.49269601702690125, "step": 355, "tools/generated_tokens": 2692.044921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.669921875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.380859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1976.0, "completions/mean_length": 1391.71484375, "completions/mean_terminated_length": 990.1892700195312, "completions/min_length": 136.0, "completions/min_terminated_length": 136.0, "epoch": 0.2526614620298084, "frac_reward_zero_std": 0.28125, "grad_norm": 7.497640609741211, "learning_rate": 2.9e-07, "loss": 0.0711, "num_tokens": 823208616.0, "reward": 0.3984375, "reward_std": 0.2737070918083191, "rewards/simpleverify_reward/mean": 0.3984375, "rewards/simpleverify_reward/std": 0.4900552034378052, "step": 356, "tools/generated_tokens": 2401.21875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.41796875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1979.0, "completions/mean_length": 1506.48828125, "completions/mean_terminated_length": 1117.617431640625, "completions/min_length": 39.0, "completions/min_terminated_length": 39.0, "epoch": 0.25337118523775726, "frac_reward_zero_std": 0.3125, "grad_norm": 4.333558082580566, "learning_rate": 2.88e-07, "loss": 0.0655, "num_tokens": 825539415.0, "reward": 0.34765625, "reward_std": 0.30011945962905884, "rewards/simpleverify_reward/mean": 0.34765625, "rewards/simpleverify_reward/std": 0.47669193148612976, "step": 357, "tools/generated_tokens": 2813.728515625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7109375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1951.0, "completions/mean_length": 1532.685546875, "completions/mean_terminated_length": 1168.530029296875, "completions/min_length": 98.0, "completions/min_terminated_length": 98.0, "epoch": 0.2540809084457062, "frac_reward_zero_std": 0.15625, "grad_norm": 7.364162921905518, "learning_rate": 2.8599999999999994e-07, "loss": 0.0476, "num_tokens": 827831151.0, "reward": 0.369140625, "reward_std": 0.32763540744781494, "rewards/simpleverify_reward/mean": 0.369140625, "rewards/simpleverify_reward/std": 0.4830440282821655, "step": 358, "tools/generated_tokens": 2769.533203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.673828125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1955.0, "completions/mean_length": 1391.2421875, "completions/mean_terminated_length": 1117.005615234375, "completions/min_length": 135.0, "completions/min_terminated_length": 135.0, "epoch": 0.25479063165365506, "frac_reward_zero_std": 0.28125, "grad_norm": 1.8164525032043457, "learning_rate": 2.8399999999999995e-07, "loss": 0.0836, "num_tokens": 829799781.0, "reward": 0.498046875, "reward_std": 0.30640435218811035, "rewards/simpleverify_reward/mean": 0.498046875, "rewards/simpleverify_reward/std": 0.5004851818084717, "step": 359, "tools/generated_tokens": 2312.96875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5078125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.357421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1985.0, "completions/mean_length": 1362.734375, "completions/mean_terminated_length": 981.5684204101562, "completions/min_length": 69.0, "completions/min_terminated_length": 69.0, "epoch": 0.255500354861604, "frac_reward_zero_std": 0.3125, "grad_norm": 4.322812557220459, "learning_rate": 2.8199999999999996e-07, "loss": 0.0332, "num_tokens": 831833556.0, "reward": 0.28125, "reward_std": 0.28014805912971497, "rewards/simpleverify_reward/mean": 0.28125, "rewards/simpleverify_reward/std": 0.45004892349243164, "step": 360, "tools/generated_tokens": 2404.041015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.56640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.31640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1940.0, "completions/mean_length": 1345.06640625, "completions/mean_terminated_length": 1019.7085571289062, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.25621007806955287, "frac_reward_zero_std": 0.28125, "grad_norm": 4.161472797393799, "learning_rate": 2.8e-07, "loss": 0.0701, "num_tokens": 833738581.0, "reward": 0.4375, "reward_std": 0.279762327671051, "rewards/simpleverify_reward/mean": 0.4375, "rewards/simpleverify_reward/std": 0.49656352400779724, "step": 361, "tools/generated_tokens": 2221.595703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.478515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.32421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2033.0, "completions/mean_length": 1379.4453125, "completions/mean_terminated_length": 1059.8497314453125, "completions/min_length": 92.0, "completions/min_terminated_length": 92.0, "epoch": 0.25691980127750175, "frac_reward_zero_std": 0.25, "grad_norm": 2.200852632522583, "learning_rate": 2.7800000000000003e-07, "loss": 0.0807, "num_tokens": 835774795.0, "reward": 0.4453125, "reward_std": 0.3109981119632721, "rewards/simpleverify_reward/mean": 0.4453125, "rewards/simpleverify_reward/std": 0.49748632311820984, "step": 362, "tools/generated_tokens": 2399.658203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.55859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.322265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1995.0, "completions/mean_length": 1357.015625, "completions/mean_terminated_length": 1028.4495849609375, "completions/min_length": 68.0, "completions/min_terminated_length": 68.0, "epoch": 0.2576295244854507, "frac_reward_zero_std": 0.21875, "grad_norm": 2.9005472660064697, "learning_rate": 2.7600000000000004e-07, "loss": 0.0528, "num_tokens": 837853887.0, "reward": 0.404296875, "reward_std": 0.3291594386100769, "rewards/simpleverify_reward/mean": 0.404296875, "rewards/simpleverify_reward/std": 0.4912354052066803, "step": 363, "tools/generated_tokens": 2451.001953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.59765625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.330078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1976.0, "completions/mean_length": 1391.39453125, "completions/mean_terminated_length": 1067.8775634765625, "completions/min_length": 94.0, "completions/min_terminated_length": 94.0, "epoch": 0.25833924769339955, "frac_reward_zero_std": 0.15625, "grad_norm": 5.4356184005737305, "learning_rate": 2.74e-07, "loss": 0.0422, "num_tokens": 839955067.0, "reward": 0.359375, "reward_std": 0.3401453197002411, "rewards/simpleverify_reward/mean": 0.359375, "rewards/simpleverify_reward/std": 0.48028653860092163, "step": 364, "tools/generated_tokens": 2464.646484375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.283203125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1947.0, "completions/mean_length": 1305.009765625, "completions/mean_terminated_length": 1011.9536743164062, "completions/min_length": 81.0, "completions/min_terminated_length": 81.0, "epoch": 0.2590489709013485, "frac_reward_zero_std": 0.3125, "grad_norm": 2.7970902919769287, "learning_rate": 2.72e-07, "loss": 0.0613, "num_tokens": 841993547.0, "reward": 0.423828125, "reward_std": 0.2828308343887329, "rewards/simpleverify_reward/mean": 0.423828125, "rewards/simpleverify_reward/std": 0.4946470856666565, "step": 365, "tools/generated_tokens": 2385.427734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.587890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1960.0, "completions/mean_length": 1345.359375, "completions/mean_terminated_length": 1091.2127685546875, "completions/min_length": 1.0, "completions/min_terminated_length": 1.0, "epoch": 0.25975869410929736, "frac_reward_zero_std": 0.1875, "grad_norm": 3.8094046115875244, "learning_rate": 2.7e-07, "loss": 0.0353, "num_tokens": 843938800.0, "reward": 0.53515625, "reward_std": 0.3339580297470093, "rewards/simpleverify_reward/mean": 0.53515625, "rewards/simpleverify_reward/std": 0.49925029277801514, "step": 366, "tools/generated_tokens": 2288.724609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.51171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1987.0, "completions/mean_length": 1467.830078125, "completions/mean_terminated_length": 1143.320068359375, "completions/min_length": 36.0, "completions/min_terminated_length": 36.0, "epoch": 0.2604684173172463, "frac_reward_zero_std": 0.3125, "grad_norm": 2.967895030975342, "learning_rate": 2.68e-07, "loss": 0.0428, "num_tokens": 846248503.0, "reward": 0.416015625, "reward_std": 0.30711260437965393, "rewards/simpleverify_reward/mean": 0.416015625, "rewards/simpleverify_reward/std": 0.493378221988678, "step": 367, "tools/generated_tokens": 2776.275390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.70703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.494140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1964.0, "completions/mean_length": 1512.517578125, "completions/mean_terminated_length": 989.4401245117188, "completions/min_length": 134.0, "completions/min_terminated_length": 134.0, "epoch": 0.26117814052519517, "frac_reward_zero_std": 0.25, "grad_norm": 2.949075222015381, "learning_rate": 2.66e-07, "loss": 0.1001, "num_tokens": 848596336.0, "reward": 0.306640625, "reward_std": 0.3109050989151001, "rewards/simpleverify_reward/mean": 0.306640625, "rewards/simpleverify_reward/std": 0.4615498185157776, "step": 368, "tools/generated_tokens": 2818.36328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.720703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.365234375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1932.0, "completions/mean_length": 1504.369140625, "completions/mean_terminated_length": 1193.0677490234375, "completions/min_length": 99.0, "completions/min_terminated_length": 99.0, "epoch": 0.2618878637331441, "frac_reward_zero_std": 0.21875, "grad_norm": 2.929680824279785, "learning_rate": 2.64e-07, "loss": 0.0647, "num_tokens": 850889384.0, "reward": 0.431640625, "reward_std": 0.3090674877166748, "rewards/simpleverify_reward/mean": 0.431640625, "rewards/simpleverify_reward/std": 0.4957893490791321, "step": 369, "tools/generated_tokens": 2755.796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.685546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2026.0, "completions/mean_length": 1463.69140625, "completions/mean_terminated_length": 994.5985717773438, "completions/min_length": 76.0, "completions/min_terminated_length": 76.0, "epoch": 0.262597586941093, "frac_reward_zero_std": 0.34375, "grad_norm": 2.3162057399749756, "learning_rate": 2.62e-07, "loss": 0.0706, "num_tokens": 853240377.0, "reward": 0.3046875, "reward_std": 0.26425856351852417, "rewards/simpleverify_reward/mean": 0.3046875, "rewards/simpleverify_reward/std": 0.4607250988483429, "step": 370, "tools/generated_tokens": 2793.5546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.724609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.244140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2009.0, "completions/mean_length": 1206.353515625, "completions/mean_terminated_length": 935.2144775390625, "completions/min_length": 149.0, "completions/min_terminated_length": 149.0, "epoch": 0.26330731014904185, "frac_reward_zero_std": 0.28125, "grad_norm": 1.7654273509979248, "learning_rate": 2.6e-07, "loss": 0.0503, "num_tokens": 854916200.0, "reward": 0.484375, "reward_std": 0.3081853985786438, "rewards/simpleverify_reward/mean": 0.484375, "rewards/simpleverify_reward/std": 0.5002445578575134, "step": 371, "tools/generated_tokens": 1896.474609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.37890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.31640625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1979.0, "completions/mean_length": 1387.412109375, "completions/mean_terminated_length": 1081.9228515625, "completions/min_length": 214.0, "completions/min_terminated_length": 214.0, "epoch": 0.2640170333569908, "frac_reward_zero_std": 0.25, "grad_norm": 3.303720235824585, "learning_rate": 2.58e-07, "loss": 0.0779, "num_tokens": 857025116.0, "reward": 0.419921875, "reward_std": 0.33301395177841187, "rewards/simpleverify_reward/mean": 0.419921875, "rewards/simpleverify_reward/std": 0.4940285086631775, "step": 372, "tools/generated_tokens": 2488.712890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.595703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.333984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1951.0, "completions/mean_length": 1389.041015625, "completions/mean_terminated_length": 1060.284423828125, "completions/min_length": 97.0, "completions/min_terminated_length": 97.0, "epoch": 0.26472675656493966, "frac_reward_zero_std": 0.21875, "grad_norm": 2.562554359436035, "learning_rate": 2.56e-07, "loss": 0.0605, "num_tokens": 859138012.0, "reward": 0.427734375, "reward_std": 0.3179839849472046, "rewards/simpleverify_reward/mean": 0.427734375, "rewards/simpleverify_reward/std": 0.4952339828014374, "step": 373, "tools/generated_tokens": 2470.609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.298828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1990.0, "completions/mean_length": 1328.99609375, "completions/mean_terminated_length": 1022.8356323242188, "completions/min_length": 163.0, "completions/min_terminated_length": 163.0, "epoch": 0.2654364797728886, "frac_reward_zero_std": 0.25, "grad_norm": 5.9632649421691895, "learning_rate": 2.5399999999999997e-07, "loss": 0.102, "num_tokens": 861116100.0, "reward": 0.482421875, "reward_std": 0.31897789239883423, "rewards/simpleverify_reward/mean": 0.482421875, "rewards/simpleverify_reward/std": 0.5001795887947083, "step": 374, "tools/generated_tokens": 2312.05859375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.53515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.423828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2006.0, "completions/mean_length": 1505.31640625, "completions/mean_terminated_length": 1107.203369140625, "completions/min_length": 50.0, "completions/min_terminated_length": 50.0, "epoch": 0.26614620298083747, "frac_reward_zero_std": 0.25, "grad_norm": 1.6678450107574463, "learning_rate": 2.52e-07, "loss": 0.0577, "num_tokens": 863451268.0, "reward": 0.39453125, "reward_std": 0.2893025875091553, "rewards/simpleverify_reward/mean": 0.39453125, "rewards/simpleverify_reward/std": 0.4892277717590332, "step": 375, "tools/generated_tokens": 2816.125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.71484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.2734375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2032.0, "completions/mean_length": 1334.279296875, "completions/mean_terminated_length": 1068.0887451171875, "completions/min_length": 52.0, "completions/min_terminated_length": 52.0, "epoch": 0.2668559261887864, "frac_reward_zero_std": 0.34375, "grad_norm": 1.9628061056137085, "learning_rate": 2.5e-07, "loss": 0.093, "num_tokens": 865428185.0, "reward": 0.470703125, "reward_std": 0.2601694166660309, "rewards/simpleverify_reward/mean": 0.470703125, "rewards/simpleverify_reward/std": 0.49962911009788513, "step": 376, "tools/generated_tokens": 2315.48046875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.53125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.373046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1970.0, "completions/mean_length": 1443.0, "completions/mean_terminated_length": 1083.0155029296875, "completions/min_length": 70.0, "completions/min_terminated_length": 70.0, "epoch": 0.2675656493967353, "frac_reward_zero_std": 0.34375, "grad_norm": 1.8627089262008667, "learning_rate": 2.48e-07, "loss": 0.0564, "num_tokens": 867599050.0, "reward": 0.357421875, "reward_std": 0.2773873805999756, "rewards/simpleverify_reward/mean": 0.357421875, "rewards/simpleverify_reward/std": 0.4797092080116272, "step": 377, "tools/generated_tokens": 2583.671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.619140625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.443359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1912.0, "completions/mean_length": 1541.041015625, "completions/mean_terminated_length": 1137.252685546875, "completions/min_length": 119.0, "completions/min_terminated_length": 119.0, "epoch": 0.26827537260468415, "frac_reward_zero_std": 0.375, "grad_norm": 2.009875774383545, "learning_rate": 2.46e-07, "loss": 0.0565, "num_tokens": 869856443.0, "reward": 0.404296875, "reward_std": 0.249843031167984, "rewards/simpleverify_reward/mean": 0.404296875, "rewards/simpleverify_reward/std": 0.4912354052066803, "step": 378, "tools/generated_tokens": 2729.986328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.650390625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1953.0, "completions/mean_length": 1293.150390625, "completions/mean_terminated_length": 962.5534057617188, "completions/min_length": 136.0, "completions/min_terminated_length": 136.0, "epoch": 0.2689850958126331, "frac_reward_zero_std": 0.28125, "grad_norm": 2.9453794956207275, "learning_rate": 2.4399999999999996e-07, "loss": 0.0982, "num_tokens": 871859224.0, "reward": 0.443359375, "reward_std": 0.3087902069091797, "rewards/simpleverify_reward/mean": 0.443359375, "rewards/simpleverify_reward/std": 0.49726733565330505, "step": 379, "tools/generated_tokens": 2337.10546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.568359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1925.0, "completions/mean_length": 1340.482421875, "completions/mean_terminated_length": 957.5842895507812, "completions/min_length": 139.0, "completions/min_terminated_length": 139.0, "epoch": 0.26969481902058196, "frac_reward_zero_std": 0.21875, "grad_norm": 1.7662068605422974, "learning_rate": 2.4199999999999997e-07, "loss": 0.0644, "num_tokens": 873893464.0, "reward": 0.439453125, "reward_std": 0.31089818477630615, "rewards/simpleverify_reward/mean": 0.439453125, "rewards/simpleverify_reward/std": 0.49680593609809875, "step": 380, "tools/generated_tokens": 2387.6640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.396484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1952.0, "completions/mean_length": 1458.2265625, "completions/mean_terminated_length": 1071.317138671875, "completions/min_length": 76.0, "completions/min_terminated_length": 76.0, "epoch": 0.2704045422285309, "frac_reward_zero_std": 0.34375, "grad_norm": 1.7037378549575806, "learning_rate": 2.4e-07, "loss": 0.0161, "num_tokens": 876065755.0, "reward": 0.404296875, "reward_std": 0.22524136304855347, "rewards/simpleverify_reward/mean": 0.404296875, "rewards/simpleverify_reward/std": 0.4912354052066803, "step": 381, "tools/generated_tokens": 2581.8359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.61328125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.32421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1967.0, "completions/mean_length": 1369.06640625, "completions/mean_terminated_length": 1045.658935546875, "completions/min_length": 117.0, "completions/min_terminated_length": 117.0, "epoch": 0.27111426543647976, "frac_reward_zero_std": 0.25, "grad_norm": 1.683225393295288, "learning_rate": 2.38e-07, "loss": 0.0692, "num_tokens": 878088058.0, "reward": 0.390625, "reward_std": 0.2903290390968323, "rewards/simpleverify_reward/mean": 0.390625, "rewards/simpleverify_reward/std": 0.48836761713027954, "step": 382, "tools/generated_tokens": 2391.458984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.517578125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2002.0, "completions/mean_length": 1582.51953125, "completions/mean_terminated_length": 1086.17822265625, "completions/min_length": 172.0, "completions/min_terminated_length": 172.0, "epoch": 0.2718239886444287, "frac_reward_zero_std": 0.3125, "grad_norm": 8.72686767578125, "learning_rate": 2.3599999999999997e-07, "loss": 0.0368, "num_tokens": 880469099.0, "reward": 0.27734375, "reward_std": 0.26622071862220764, "rewards/simpleverify_reward/mean": 0.27734375, "rewards/simpleverify_reward/std": 0.4481254518032074, "step": 383, "tools/generated_tokens": 2881.255859375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.716796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2007.0, "completions/mean_length": 1534.599609375, "completions/mean_terminated_length": 1216.161376953125, "completions/min_length": 43.0, "completions/min_terminated_length": 43.0, "epoch": 0.27253371185237757, "frac_reward_zero_std": 0.21875, "grad_norm": 3.8800652027130127, "learning_rate": 2.34e-07, "loss": 0.0571, "num_tokens": 882805203.0, "reward": 0.396484375, "reward_std": 0.3169543743133545, "rewards/simpleverify_reward/mean": 0.396484375, "rewards/simpleverify_reward/std": 0.4896455705165863, "step": 384, "tools/generated_tokens": 2834.5546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.326171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1996.0, "completions/mean_length": 1409.4921875, "completions/mean_terminated_length": 1100.5797119140625, "completions/min_length": 240.0, "completions/min_terminated_length": 240.0, "epoch": 0.27324343506032645, "frac_reward_zero_std": 0.28125, "grad_norm": 2.095346689224243, "learning_rate": 2.32e-07, "loss": 0.0381, "num_tokens": 884874070.0, "reward": 0.46875, "reward_std": 0.27535808086395264, "rewards/simpleverify_reward/mean": 0.46875, "rewards/simpleverify_reward/std": 0.4995105266571045, "step": 385, "tools/generated_tokens": 2446.603515625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.564453125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.37109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2002.0, "completions/mean_length": 1400.748046875, "completions/mean_terminated_length": 1018.8292236328125, "completions/min_length": 32.0, "completions/min_terminated_length": 32.0, "epoch": 0.2739531582682754, "frac_reward_zero_std": 0.1875, "grad_norm": 3.805717706680298, "learning_rate": 2.3e-07, "loss": 0.0573, "num_tokens": 887075828.0, "reward": 0.384765625, "reward_std": 0.31686756014823914, "rewards/simpleverify_reward/mean": 0.384765625, "rewards/simpleverify_reward/std": 0.4870156943798065, "step": 386, "tools/generated_tokens": 2575.455078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.638671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.357421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1962.0, "completions/mean_length": 1445.919921875, "completions/mean_terminated_length": 1111.0242919921875, "completions/min_length": 108.0, "completions/min_terminated_length": 108.0, "epoch": 0.27466288147622425, "frac_reward_zero_std": 0.34375, "grad_norm": 1.6014423370361328, "learning_rate": 2.28e-07, "loss": 0.0542, "num_tokens": 889258543.0, "reward": 0.42578125, "reward_std": 0.25226327776908875, "rewards/simpleverify_reward/mean": 0.42578125, "rewards/simpleverify_reward/std": 0.4949444830417633, "step": 387, "tools/generated_tokens": 2612.376953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.634765625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.341796875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1943.0, "completions/mean_length": 1419.890625, "completions/mean_terminated_length": 1095.824951171875, "completions/min_length": 187.0, "completions/min_terminated_length": 187.0, "epoch": 0.2753726046841732, "frac_reward_zero_std": 0.25, "grad_norm": 3.3459980487823486, "learning_rate": 2.2599999999999999e-07, "loss": 0.055, "num_tokens": 891347002.0, "reward": 0.419921875, "reward_std": 0.2843603491783142, "rewards/simpleverify_reward/mean": 0.419921875, "rewards/simpleverify_reward/std": 0.4940285086631775, "step": 388, "tools/generated_tokens": 2494.1796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.583984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.42578125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1943.0, "completions/mean_length": 1492.294921875, "completions/mean_terminated_length": 1080.241455078125, "completions/min_length": 105.0, "completions/min_terminated_length": 105.0, "epoch": 0.27608232789212206, "frac_reward_zero_std": 0.3125, "grad_norm": 5.454160213470459, "learning_rate": 2.24e-07, "loss": 0.0476, "num_tokens": 893609345.0, "reward": 0.34765625, "reward_std": 0.2748687267303467, "rewards/simpleverify_reward/mean": 0.34765625, "rewards/simpleverify_reward/std": 0.47669193148612976, "step": 389, "tools/generated_tokens": 2723.3359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.38671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1948.0, "completions/mean_length": 1392.166015625, "completions/mean_terminated_length": 981.289794921875, "completions/min_length": 159.0, "completions/min_terminated_length": 159.0, "epoch": 0.276792051100071, "frac_reward_zero_std": 0.375, "grad_norm": 1.5564805269241333, "learning_rate": 2.22e-07, "loss": 0.0287, "num_tokens": 895652885.0, "reward": 0.333984375, "reward_std": 0.2363744080066681, "rewards/simpleverify_reward/mean": 0.333984375, "rewards/simpleverify_reward/std": 0.47209542989730835, "step": 390, "tools/generated_tokens": 2410.22265625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.443359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1888.0, "completions/mean_length": 1502.298828125, "completions/mean_terminated_length": 1071.0245361328125, "completions/min_length": 49.0, "completions/min_terminated_length": 49.0, "epoch": 0.27750177430801987, "frac_reward_zero_std": 0.25, "grad_norm": 2.516817808151245, "learning_rate": 2.1999999999999998e-07, "loss": 0.0834, "num_tokens": 897908507.0, "reward": 0.451171875, "reward_std": 0.2775598168373108, "rewards/simpleverify_reward/mean": 0.451171875, "rewards/simpleverify_reward/std": 0.498096764087677, "step": 391, "tools/generated_tokens": 2698.826171875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.65234375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.361328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1970.0, "completions/mean_length": 1426.28125, "completions/mean_terminated_length": 1075.865478515625, "completions/min_length": 43.0, "completions/min_terminated_length": 43.0, "epoch": 0.2782114975159688, "frac_reward_zero_std": 0.25, "grad_norm": 2.437361240386963, "learning_rate": 2.18e-07, "loss": 0.0559, "num_tokens": 899942248.0, "reward": 0.390625, "reward_std": 0.2886054515838623, "rewards/simpleverify_reward/mean": 0.390625, "rewards/simpleverify_reward/std": 0.48836761713027954, "step": 392, "tools/generated_tokens": 2407.267578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.53515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.396484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2012.0, "completions/mean_length": 1428.05078125, "completions/mean_terminated_length": 1021.7217407226562, "completions/min_length": 222.0, "completions/min_terminated_length": 222.0, "epoch": 0.2789212207239177, "frac_reward_zero_std": 0.3125, "grad_norm": 3.347823143005371, "learning_rate": 2.1599999999999998e-07, "loss": 0.051, "num_tokens": 902136261.0, "reward": 0.443359375, "reward_std": 0.2854210436344147, "rewards/simpleverify_reward/mean": 0.443359375, "rewards/simpleverify_reward/std": 0.49726733565330505, "step": 393, "tools/generated_tokens": 2618.201171875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.646484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.443359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1926.0, "completions/mean_length": 1494.111328125, "completions/mean_terminated_length": 1052.94384765625, "completions/min_length": 55.0, "completions/min_terminated_length": 55.0, "epoch": 0.27963094393186655, "frac_reward_zero_std": 0.25, "grad_norm": 3.383227825164795, "learning_rate": 2.1399999999999998e-07, "loss": 0.0672, "num_tokens": 904405077.0, "reward": 0.33203125, "reward_std": 0.3103010058403015, "rewards/simpleverify_reward/mean": 0.33203125, "rewards/simpleverify_reward/std": 0.47140273451805115, "step": 394, "tools/generated_tokens": 2734.337890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.67578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1984.0, "completions/mean_length": 1398.08203125, "completions/mean_terminated_length": 938.8067016601562, "completions/min_length": 22.0, "completions/min_terminated_length": 22.0, "epoch": 0.2803406671398155, "frac_reward_zero_std": 0.3125, "grad_norm": 2.145132303237915, "learning_rate": 2.12e-07, "loss": 0.0443, "num_tokens": 906531654.0, "reward": 0.310546875, "reward_std": 0.29493510723114014, "rewards/simpleverify_reward/mean": 0.310546875, "rewards/simpleverify_reward/std": 0.46317005157470703, "step": 395, "tools/generated_tokens": 2507.59765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.607421875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.205078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1989.0, "completions/mean_length": 1249.265625, "completions/mean_terminated_length": 1043.4791259765625, "completions/min_length": 117.0, "completions/min_terminated_length": 117.0, "epoch": 0.28105039034776436, "frac_reward_zero_std": 0.25, "grad_norm": 8.402230262756348, "learning_rate": 2.0999999999999997e-07, "loss": 0.03, "num_tokens": 908274282.0, "reward": 0.490234375, "reward_std": 0.30637410283088684, "rewards/simpleverify_reward/mean": 0.490234375, "rewards/simpleverify_reward/std": 0.5003935098648071, "step": 396, "tools/generated_tokens": 1993.333984375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.41015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.392578125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1984.0, "completions/mean_length": 1497.330078125, "completions/mean_terminated_length": 1141.430908203125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.2817601135557133, "frac_reward_zero_std": 0.28125, "grad_norm": 3.514080047607422, "learning_rate": 2.0799999999999998e-07, "loss": 0.0677, "num_tokens": 910514774.0, "reward": 0.404296875, "reward_std": 0.30407148599624634, "rewards/simpleverify_reward/mean": 0.404296875, "rewards/simpleverify_reward/std": 0.4912354052066803, "step": 397, "tools/generated_tokens": 2673.3203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.642578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1964.0, "completions/mean_length": 1451.236328125, "completions/mean_terminated_length": 1018.0405883789062, "completions/min_length": 1.0, "completions/min_terminated_length": 1.0, "epoch": 0.28246983676366216, "frac_reward_zero_std": 0.28125, "grad_norm": 3.9625115394592285, "learning_rate": 2.06e-07, "loss": 0.0806, "num_tokens": 912663547.0, "reward": 0.400390625, "reward_std": 0.2793550193309784, "rewards/simpleverify_reward/mean": 0.400390625, "rewards/simpleverify_reward/std": 0.4904567301273346, "step": 398, "tools/generated_tokens": 2569.55078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.61328125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.318359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2009.0, "completions/mean_length": 1470.24609375, "completions/mean_terminated_length": 1201.2149658203125, "completions/min_length": 69.0, "completions/min_terminated_length": 69.0, "epoch": 0.2831795599716111, "frac_reward_zero_std": 0.1875, "grad_norm": 2.6517229080200195, "learning_rate": 2.0399999999999997e-07, "loss": 0.0946, "num_tokens": 914907269.0, "reward": 0.373046875, "reward_std": 0.34551775455474854, "rewards/simpleverify_reward/mean": 0.373046875, "rewards/simpleverify_reward/std": 0.48408737778663635, "step": 399, "tools/generated_tokens": 2708.18359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.478515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1971.0, "completions/mean_length": 1536.4609375, "completions/mean_terminated_length": 1067.0711669921875, "completions/min_length": 204.0, "completions/min_terminated_length": 204.0, "epoch": 0.28388928317955997, "frac_reward_zero_std": 0.46875, "grad_norm": 1.087140440940857, "learning_rate": 2.02e-07, "loss": 0.0365, "num_tokens": 917242471.0, "reward": 0.24609375, "reward_std": 0.18657684326171875, "rewards/simpleverify_reward/mean": 0.24609375, "rewards/simpleverify_reward/std": 0.4311550557613373, "step": 400, "tools/generated_tokens": 2821.908203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.408203125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1964.0, "completions/mean_length": 1463.83203125, "completions/mean_terminated_length": 1061.234375, "completions/min_length": 82.0, "completions/min_terminated_length": 82.0, "epoch": 0.28459900638750885, "frac_reward_zero_std": 0.21875, "grad_norm": 2.6030118465423584, "learning_rate": 2e-07, "loss": 0.0872, "num_tokens": 919494151.0, "reward": 0.296875, "reward_std": 0.3137928545475006, "rewards/simpleverify_reward/mean": 0.296875, "rewards/simpleverify_reward/std": 0.45732781291007996, "step": 401, "tools/generated_tokens": 2703.10546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.400390625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1962.0, "completions/mean_length": 1518.970703125, "completions/mean_terminated_length": 1165.7100830078125, "completions/min_length": 140.0, "completions/min_terminated_length": 140.0, "epoch": 0.2853087295954578, "frac_reward_zero_std": 0.375, "grad_norm": 2.3760340213775635, "learning_rate": 1.98e-07, "loss": 0.079, "num_tokens": 921780303.0, "reward": 0.283203125, "reward_std": 0.259396493434906, "rewards/simpleverify_reward/mean": 0.283203125, "rewards/simpleverify_reward/std": 0.4509948492050171, "step": 402, "tools/generated_tokens": 2749.68359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.30859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1997.0, "completions/mean_length": 1409.70703125, "completions/mean_terminated_length": 1127.80224609375, "completions/min_length": 145.0, "completions/min_terminated_length": 145.0, "epoch": 0.28601845280340665, "frac_reward_zero_std": 0.3125, "grad_norm": 2.6638870239257812, "learning_rate": 1.96e-07, "loss": 0.047, "num_tokens": 923804112.0, "reward": 0.357421875, "reward_std": 0.2759947180747986, "rewards/simpleverify_reward/mean": 0.357421875, "rewards/simpleverify_reward/std": 0.4797092080116272, "step": 403, "tools/generated_tokens": 2389.3359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.533203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1976.0, "completions/mean_length": 1374.666015625, "completions/mean_terminated_length": 1034.11767578125, "completions/min_length": 182.0, "completions/min_terminated_length": 182.0, "epoch": 0.2867281760113556, "frac_reward_zero_std": 0.25, "grad_norm": 7.677732944488525, "learning_rate": 1.94e-07, "loss": 0.0399, "num_tokens": 925844009.0, "reward": 0.43359375, "reward_std": 0.31551212072372437, "rewards/simpleverify_reward/mean": 0.43359375, "rewards/simpleverify_reward/std": 0.4960552453994751, "step": 404, "tools/generated_tokens": 2393.029296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.560546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.2890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2003.0, "completions/mean_length": 1369.408203125, "completions/mean_terminated_length": 1094.445068359375, "completions/min_length": 277.0, "completions/min_terminated_length": 277.0, "epoch": 0.28743789921930446, "frac_reward_zero_std": 0.21875, "grad_norm": 4.190894603729248, "learning_rate": 1.92e-07, "loss": 0.0952, "num_tokens": 927909505.0, "reward": 0.431640625, "reward_std": 0.34279412031173706, "rewards/simpleverify_reward/mean": 0.431640625, "rewards/simpleverify_reward/std": 0.4957893490791321, "step": 405, "tools/generated_tokens": 2430.5625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.572265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.42578125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1959.0, "completions/mean_length": 1552.923828125, "completions/mean_terminated_length": 1185.9217529296875, "completions/min_length": 139.0, "completions/min_terminated_length": 139.0, "epoch": 0.2881476224272534, "frac_reward_zero_std": 0.28125, "grad_norm": 4.215174674987793, "learning_rate": 1.8999999999999998e-07, "loss": 0.0627, "num_tokens": 930240295.0, "reward": 0.28125, "reward_std": 0.2700323164463043, "rewards/simpleverify_reward/mean": 0.28125, "rewards/simpleverify_reward/std": 0.45004892349243164, "step": 406, "tools/generated_tokens": 2834.19921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.48046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1962.0, "completions/mean_length": 1551.30859375, "completions/mean_terminated_length": 1091.96240234375, "completions/min_length": 72.0, "completions/min_terminated_length": 72.0, "epoch": 0.28885734563520227, "frac_reward_zero_std": 0.21875, "grad_norm": 4.237377643585205, "learning_rate": 1.88e-07, "loss": 0.0636, "num_tokens": 932570440.0, "reward": 0.318359375, "reward_std": 0.32135486602783203, "rewards/simpleverify_reward/mean": 0.318359375, "rewards/simpleverify_reward/std": 0.46629536151885986, "step": 407, "tools/generated_tokens": 2820.609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.69140625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1899.0, "completions/mean_length": 1330.046875, "completions/mean_terminated_length": 1071.5531005859375, "completions/min_length": 81.0, "completions/min_terminated_length": 81.0, "epoch": 0.2895670688431512, "frac_reward_zero_std": 0.1875, "grad_norm": 2.7384283542633057, "learning_rate": 1.86e-07, "loss": 0.0222, "num_tokens": 934508384.0, "reward": 0.427734375, "reward_std": 0.3449370563030243, "rewards/simpleverify_reward/mean": 0.427734375, "rewards/simpleverify_reward/std": 0.4952339828014374, "step": 408, "tools/generated_tokens": 2255.5703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.501953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.396484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1934.0, "completions/mean_length": 1410.337890625, "completions/mean_terminated_length": 991.4207153320312, "completions/min_length": 40.0, "completions/min_terminated_length": 40.0, "epoch": 0.2902767920511001, "frac_reward_zero_std": 0.28125, "grad_norm": 1.8199025392532349, "learning_rate": 1.8399999999999998e-07, "loss": 0.0537, "num_tokens": 936699354.0, "reward": 0.263671875, "reward_std": 0.2761150002479553, "rewards/simpleverify_reward/mean": 0.263671875, "rewards/simpleverify_reward/std": 0.4410543739795685, "step": 409, "tools/generated_tokens": 2592.828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.30859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1959.0, "completions/mean_length": 1334.123046875, "completions/mean_terminated_length": 1015.5, "completions/min_length": 130.0, "completions/min_terminated_length": 130.0, "epoch": 0.29098651525904895, "frac_reward_zero_std": 0.375, "grad_norm": 2.0764877796173096, "learning_rate": 1.82e-07, "loss": 0.0477, "num_tokens": 938733134.0, "reward": 0.408203125, "reward_std": 0.2584351599216461, "rewards/simpleverify_reward/mean": 0.408203125, "rewards/simpleverify_reward/std": 0.49198177456855774, "step": 410, "tools/generated_tokens": 2372.34765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.564453125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.400390625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1953.0, "completions/mean_length": 1460.43359375, "completions/mean_terminated_length": 1069.048828125, "completions/min_length": 64.0, "completions/min_terminated_length": 64.0, "epoch": 0.2916962384669979, "frac_reward_zero_std": 0.34375, "grad_norm": 2.2778093814849854, "learning_rate": 1.8e-07, "loss": 0.0736, "num_tokens": 940867507.0, "reward": 0.39453125, "reward_std": 0.27205604314804077, "rewards/simpleverify_reward/mean": 0.39453125, "rewards/simpleverify_reward/std": 0.4892277717590332, "step": 411, "tools/generated_tokens": 2549.431640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.591796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.294921875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1986.0, "completions/mean_length": 1358.380859375, "completions/mean_terminated_length": 1070.97509765625, "completions/min_length": 86.0, "completions/min_terminated_length": 86.0, "epoch": 0.29240596167494676, "frac_reward_zero_std": 0.15625, "grad_norm": 2.705021381378174, "learning_rate": 1.7799999999999998e-07, "loss": 0.0685, "num_tokens": 942918041.0, "reward": 0.462890625, "reward_std": 0.377566933631897, "rewards/simpleverify_reward/mean": 0.462890625, "rewards/simpleverify_reward/std": 0.4991086423397064, "step": 412, "tools/generated_tokens": 2422.21484375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.57421875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1957.0, "completions/mean_length": 1391.484375, "completions/mean_terminated_length": 1105.07861328125, "completions/min_length": 46.0, "completions/min_terminated_length": 46.0, "epoch": 0.2931156848828957, "frac_reward_zero_std": 0.3125, "grad_norm": 3.1275455951690674, "learning_rate": 1.76e-07, "loss": 0.0294, "num_tokens": 944970031.0, "reward": 0.392578125, "reward_std": 0.2591443359851837, "rewards/simpleverify_reward/mean": 0.392578125, "rewards/simpleverify_reward/std": 0.4888018071651459, "step": 413, "tools/generated_tokens": 2429.857421875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.56640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.34765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1989.0, "completions/mean_length": 1376.515625, "completions/mean_terminated_length": 1019.3712768554688, "completions/min_length": 169.0, "completions/min_terminated_length": 169.0, "epoch": 0.29382540809084456, "frac_reward_zero_std": 0.28125, "grad_norm": 4.306106090545654, "learning_rate": 1.7399999999999997e-07, "loss": 0.0978, "num_tokens": 947054173.0, "reward": 0.41796875, "reward_std": 0.2804032564163208, "rewards/simpleverify_reward/mean": 0.41796875, "rewards/simpleverify_reward/std": 0.4937073290348053, "step": 414, "tools/generated_tokens": 2430.078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.271484375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1950.0, "completions/mean_length": 1334.916015625, "completions/mean_terminated_length": 1069.4959716796875, "completions/min_length": 65.0, "completions/min_terminated_length": 65.0, "epoch": 0.2945351312987935, "frac_reward_zero_std": 0.1875, "grad_norm": 4.8833231925964355, "learning_rate": 1.7199999999999998e-07, "loss": 0.065, "num_tokens": 949015322.0, "reward": 0.50390625, "reward_std": 0.3353680968284607, "rewards/simpleverify_reward/mean": 0.50390625, "rewards/simpleverify_reward/std": 0.5004737377166748, "step": 415, "tools/generated_tokens": 2285.015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.384765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1977.0, "completions/mean_length": 1463.998046875, "completions/mean_terminated_length": 1100.869873046875, "completions/min_length": 250.0, "completions/min_terminated_length": 250.0, "epoch": 0.29524485450674237, "frac_reward_zero_std": 0.3125, "grad_norm": 2.3165814876556396, "learning_rate": 1.7000000000000001e-07, "loss": 0.0944, "num_tokens": 951279002.0, "reward": 0.39453125, "reward_std": 0.29383042454719543, "rewards/simpleverify_reward/mean": 0.39453125, "rewards/simpleverify_reward/std": 0.4892277717590332, "step": 416, "tools/generated_tokens": 2698.564453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.669921875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.263671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2005.0, "completions/mean_length": 1260.373046875, "completions/mean_terminated_length": 979.392578125, "completions/min_length": 159.0, "completions/min_terminated_length": 159.0, "epoch": 0.29595457771469125, "frac_reward_zero_std": 0.25, "grad_norm": 2.8964953422546387, "learning_rate": 1.68e-07, "loss": 0.0437, "num_tokens": 953047094.0, "reward": 0.451171875, "reward_std": 0.2922522723674774, "rewards/simpleverify_reward/mean": 0.451171875, "rewards/simpleverify_reward/std": 0.498096764087677, "step": 417, "tools/generated_tokens": 2038.6484375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.42578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.33203125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1961.0, "completions/mean_length": 1356.087890625, "completions/mean_terminated_length": 1015.23974609375, "completions/min_length": 157.0, "completions/min_terminated_length": 157.0, "epoch": 0.2966643009226402, "frac_reward_zero_std": 0.1875, "grad_norm": 2.6292054653167725, "learning_rate": 1.66e-07, "loss": 0.0628, "num_tokens": 955078578.0, "reward": 0.3671875, "reward_std": 0.3318679928779602, "rewards/simpleverify_reward/mean": 0.3671875, "rewards/simpleverify_reward/std": 0.48250964283943176, "step": 418, "tools/generated_tokens": 2387.29296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.560546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.4296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1994.0, "completions/mean_length": 1473.48046875, "completions/mean_terminated_length": 1044.23291015625, "completions/min_length": 154.0, "completions/min_terminated_length": 154.0, "epoch": 0.29737402413058905, "frac_reward_zero_std": 0.40625, "grad_norm": 1.7755942344665527, "learning_rate": 1.64e-07, "loss": 0.0539, "num_tokens": 957366043.0, "reward": 0.318359375, "reward_std": 0.23239317536354065, "rewards/simpleverify_reward/mean": 0.318359375, "rewards/simpleverify_reward/std": 0.46629536151885986, "step": 419, "tools/generated_tokens": 2735.66796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.68359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.369140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1915.0, "completions/mean_length": 1417.154296875, "completions/mean_terminated_length": 1048.715087890625, "completions/min_length": 124.0, "completions/min_terminated_length": 124.0, "epoch": 0.298083747338538, "frac_reward_zero_std": 0.25, "grad_norm": 3.6719233989715576, "learning_rate": 1.62e-07, "loss": 0.0504, "num_tokens": 959487061.0, "reward": 0.4140625, "reward_std": 0.29556477069854736, "rewards/simpleverify_reward/mean": 0.4140625, "rewards/simpleverify_reward/std": 0.49304109811782837, "step": 420, "tools/generated_tokens": 2528.28515625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.33203125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1992.0, "completions/mean_length": 1461.732421875, "completions/mean_terminated_length": 1170.906494140625, "completions/min_length": 171.0, "completions/min_terminated_length": 171.0, "epoch": 0.29879347054648686, "frac_reward_zero_std": 0.3125, "grad_norm": 4.246241569519043, "learning_rate": 1.6e-07, "loss": 0.0573, "num_tokens": 961662254.0, "reward": 0.421875, "reward_std": 0.29352304339408875, "rewards/simpleverify_reward/mean": 0.421875, "rewards/simpleverify_reward/std": 0.49434176087379456, "step": 421, "tools/generated_tokens": 2600.9609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.62109375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.322265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2037.0, "completions/mean_length": 1327.919921875, "completions/mean_terminated_length": 986.4754638671875, "completions/min_length": 214.0, "completions/min_terminated_length": 214.0, "epoch": 0.2995031937544358, "frac_reward_zero_std": 0.40625, "grad_norm": 1.8694629669189453, "learning_rate": 1.5799999999999999e-07, "loss": 0.0537, "num_tokens": 963591575.0, "reward": 0.376953125, "reward_std": 0.2416246235370636, "rewards/simpleverify_reward/mean": 0.376953125, "rewards/simpleverify_reward/std": 0.4850969910621643, "step": 422, "tools/generated_tokens": 2239.64453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.498046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.330078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1973.0, "completions/mean_length": 1389.609375, "completions/mean_terminated_length": 1066.3323974609375, "completions/min_length": 42.0, "completions/min_terminated_length": 42.0, "epoch": 0.30021291696238467, "frac_reward_zero_std": 0.375, "grad_norm": 4.614813327789307, "learning_rate": 1.56e-07, "loss": 0.0631, "num_tokens": 965616591.0, "reward": 0.447265625, "reward_std": 0.2766861915588379, "rewards/simpleverify_reward/mean": 0.447265625, "rewards/simpleverify_reward/std": 0.4976975917816162, "step": 423, "tools/generated_tokens": 2386.240234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.541015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.357421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1975.0, "completions/mean_length": 1448.05078125, "completions/mean_terminated_length": 1115.6170654296875, "completions/min_length": 95.0, "completions/min_terminated_length": 95.0, "epoch": 0.30092264017033354, "frac_reward_zero_std": 0.3125, "grad_norm": 4.802687644958496, "learning_rate": 1.54e-07, "loss": 0.0708, "num_tokens": 967773453.0, "reward": 0.416015625, "reward_std": 0.2907769978046417, "rewards/simpleverify_reward/mean": 0.416015625, "rewards/simpleverify_reward/std": 0.493378221988678, "step": 424, "tools/generated_tokens": 2547.6953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.607421875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.451171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2013.0, "completions/mean_length": 1600.853515625, "completions/mean_terminated_length": 1233.2705078125, "completions/min_length": 137.0, "completions/min_terminated_length": 137.0, "epoch": 0.3016323633782825, "frac_reward_zero_std": 0.25, "grad_norm": 4.202091217041016, "learning_rate": 1.5199999999999998e-07, "loss": 0.0682, "num_tokens": 970120012.0, "reward": 0.36328125, "reward_std": 0.32596421241760254, "rewards/simpleverify_reward/mean": 0.36328125, "rewards/simpleverify_reward/std": 0.4814152419567108, "step": 425, "tools/generated_tokens": 2849.314453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.68359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.373046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1960.0, "completions/mean_length": 1408.59765625, "completions/mean_terminated_length": 1028.143310546875, "completions/min_length": 41.0, "completions/min_terminated_length": 41.0, "epoch": 0.30234208658623135, "frac_reward_zero_std": 0.3125, "grad_norm": 2.6163883209228516, "learning_rate": 1.5e-07, "loss": 0.0591, "num_tokens": 972251090.0, "reward": 0.451171875, "reward_std": 0.29239240288734436, "rewards/simpleverify_reward/mean": 0.451171875, "rewards/simpleverify_reward/std": 0.498096764087677, "step": 426, "tools/generated_tokens": 2517.501953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.60546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.345703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1956.0, "completions/mean_length": 1401.466796875, "completions/mean_terminated_length": 1061.7164306640625, "completions/min_length": 75.0, "completions/min_terminated_length": 75.0, "epoch": 0.3030518097941803, "frac_reward_zero_std": 0.375, "grad_norm": 1.6780482530593872, "learning_rate": 1.4799999999999998e-07, "loss": 0.0508, "num_tokens": 974414504.0, "reward": 0.4140625, "reward_std": 0.2675633430480957, "rewards/simpleverify_reward/mean": 0.4140625, "rewards/simpleverify_reward/std": 0.49304109811782837, "step": 427, "tools/generated_tokens": 2568.921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.62890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.384765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1988.0, "completions/mean_length": 1427.8359375, "completions/mean_terminated_length": 1039.9873046875, "completions/min_length": 211.0, "completions/min_terminated_length": 211.0, "epoch": 0.30376153300212916, "frac_reward_zero_std": 0.28125, "grad_norm": 3.675152063369751, "learning_rate": 1.4599999999999998e-07, "loss": 0.0507, "num_tokens": 976573461.0, "reward": 0.47265625, "reward_std": 0.29628029465675354, "rewards/simpleverify_reward/mean": 0.47265625, "rewards/simpleverify_reward/std": 0.49974003434181213, "step": 428, "tools/generated_tokens": 2566.921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.62109375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.27734375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1946.0, "completions/mean_length": 1396.8359375, "completions/mean_terminated_length": 1148.5567626953125, "completions/min_length": 181.0, "completions/min_terminated_length": 181.0, "epoch": 0.3044712562100781, "frac_reward_zero_std": 0.25, "grad_norm": 2.913996458053589, "learning_rate": 1.44e-07, "loss": 0.057, "num_tokens": 978643830.0, "reward": 0.404296875, "reward_std": 0.3135787844657898, "rewards/simpleverify_reward/mean": 0.404296875, "rewards/simpleverify_reward/std": 0.4912354052066803, "step": 429, "tools/generated_tokens": 2462.732421875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.57421875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1966.0, "completions/mean_length": 1520.361328125, "completions/mean_terminated_length": 1039.973876953125, "completions/min_length": 83.0, "completions/min_terminated_length": 83.0, "epoch": 0.30518097941802697, "frac_reward_zero_std": 0.3125, "grad_norm": 3.3310375213623047, "learning_rate": 1.4199999999999997e-07, "loss": 0.0621, "num_tokens": 981000350.0, "reward": 0.34375, "reward_std": 0.26044297218322754, "rewards/simpleverify_reward/mean": 0.34375, "rewards/simpleverify_reward/std": 0.4754233956336975, "step": 430, "tools/generated_tokens": 2810.974609375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.705078125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.33984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1946.0, "completions/mean_length": 1402.1484375, "completions/mean_terminated_length": 1070.25146484375, "completions/min_length": 155.0, "completions/min_terminated_length": 155.0, "epoch": 0.3058907026259759, "frac_reward_zero_std": 0.25, "grad_norm": 2.3364529609680176, "learning_rate": 1.4e-07, "loss": 0.0397, "num_tokens": 983042363.0, "reward": 0.375, "reward_std": 0.29574164748191833, "rewards/simpleverify_reward/mean": 0.375, "rewards/simpleverify_reward/std": 0.4845963716506958, "step": 431, "tools/generated_tokens": 2429.61328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.55859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.39453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2008.0, "completions/mean_length": 1490.4609375, "completions/mean_terminated_length": 1128.5870361328125, "completions/min_length": 286.0, "completions/min_terminated_length": 286.0, "epoch": 0.3066004258339248, "frac_reward_zero_std": 0.28125, "grad_norm": 1.4284873008728027, "learning_rate": 1.3800000000000002e-07, "loss": 0.0363, "num_tokens": 985277274.0, "reward": 0.353515625, "reward_std": 0.30133551359176636, "rewards/simpleverify_reward/mean": 0.353515625, "rewards/simpleverify_reward/std": 0.47852855920791626, "step": 432, "tools/generated_tokens": 2677.421875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.64453125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.2734375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1957.0, "completions/mean_length": 1301.62109375, "completions/mean_terminated_length": 1022.3494873046875, "completions/min_length": 214.0, "completions/min_terminated_length": 214.0, "epoch": 0.30731014904187365, "frac_reward_zero_std": 0.34375, "grad_norm": 2.968947649002075, "learning_rate": 1.36e-07, "loss": 0.072, "num_tokens": 987181020.0, "reward": 0.564453125, "reward_std": 0.274637371301651, "rewards/simpleverify_reward/mean": 0.564453125, "rewards/simpleverify_reward/std": 0.49631330370903015, "step": 433, "tools/generated_tokens": 2215.65234375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.49609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.345703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2013.0, "completions/mean_length": 1417.068359375, "completions/mean_terminated_length": 1085.4000244140625, "completions/min_length": 28.0, "completions/min_terminated_length": 28.0, "epoch": 0.3080198722498226, "frac_reward_zero_std": 0.21875, "grad_norm": 2.7670400142669678, "learning_rate": 1.34e-07, "loss": 0.0621, "num_tokens": 989297369.0, "reward": 0.443359375, "reward_std": 0.32957959175109863, "rewards/simpleverify_reward/mean": 0.443359375, "rewards/simpleverify_reward/std": 0.49726733565330505, "step": 434, "tools/generated_tokens": 2521.244140625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.59765625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.310546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1962.0, "completions/mean_length": 1366.283203125, "completions/mean_terminated_length": 1059.220947265625, "completions/min_length": 121.0, "completions/min_terminated_length": 121.0, "epoch": 0.30872959545777146, "frac_reward_zero_std": 0.21875, "grad_norm": 1.9756923913955688, "learning_rate": 1.32e-07, "loss": 0.0383, "num_tokens": 991462324.0, "reward": 0.470703125, "reward_std": 0.3304902911186218, "rewards/simpleverify_reward/mean": 0.470703125, "rewards/simpleverify_reward/std": 0.49962911009788513, "step": 435, "tools/generated_tokens": 2552.890625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.640625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.369140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1966.0, "completions/mean_length": 1437.84375, "completions/mean_terminated_length": 1080.8172607421875, "completions/min_length": 40.0, "completions/min_terminated_length": 40.0, "epoch": 0.3094393186657204, "frac_reward_zero_std": 0.28125, "grad_norm": 2.157125473022461, "learning_rate": 1.3e-07, "loss": 0.0745, "num_tokens": 993631228.0, "reward": 0.43359375, "reward_std": 0.2759825587272644, "rewards/simpleverify_reward/mean": 0.43359375, "rewards/simpleverify_reward/std": 0.4960552453994751, "step": 436, "tools/generated_tokens": 2584.419921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.623046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.291015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1954.0, "completions/mean_length": 1330.740234375, "completions/mean_terminated_length": 1036.80712890625, "completions/min_length": 137.0, "completions/min_terminated_length": 137.0, "epoch": 0.31014904187366926, "frac_reward_zero_std": 0.125, "grad_norm": 2.780632257461548, "learning_rate": 1.28e-07, "loss": 0.0734, "num_tokens": 995680716.0, "reward": 0.376953125, "reward_std": 0.3455655574798584, "rewards/simpleverify_reward/mean": 0.376953125, "rewards/simpleverify_reward/std": 0.4850969910621643, "step": 437, "tools/generated_tokens": 2399.97265625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.580078125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1987.0, "completions/mean_length": 1505.84375, "completions/mean_terminated_length": 1084.1666259765625, "completions/min_length": 79.0, "completions/min_terminated_length": 79.0, "epoch": 0.3108587650816182, "frac_reward_zero_std": 0.15625, "grad_norm": 2.356558084487915, "learning_rate": 1.26e-07, "loss": 0.0772, "num_tokens": 997951565.0, "reward": 0.365234375, "reward_std": 0.3279300034046173, "rewards/simpleverify_reward/mean": 0.365234375, "rewards/simpleverify_reward/std": 0.4819667339324951, "step": 438, "tools/generated_tokens": 2742.619140625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.673828125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.37109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2022.0, "completions/mean_length": 1445.736328125, "completions/mean_terminated_length": 1091.7763671875, "completions/min_length": 55.0, "completions/min_terminated_length": 55.0, "epoch": 0.31156848828956707, "frac_reward_zero_std": 0.25, "grad_norm": 1.6718311309814453, "learning_rate": 1.24e-07, "loss": 0.053, "num_tokens": 1000075091.0, "reward": 0.404296875, "reward_std": 0.3225529193878174, "rewards/simpleverify_reward/mean": 0.404296875, "rewards/simpleverify_reward/std": 0.4912354052066803, "step": 439, "tools/generated_tokens": 2541.392578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.595703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.28515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1961.0, "completions/mean_length": 1252.98046875, "completions/mean_terminated_length": 936.415283203125, "completions/min_length": 70.0, "completions/min_terminated_length": 70.0, "epoch": 0.31227821149751595, "frac_reward_zero_std": 0.28125, "grad_norm": 1.8672159910202026, "learning_rate": 1.2199999999999998e-07, "loss": 0.0815, "num_tokens": 1001986288.0, "reward": 0.41015625, "reward_std": 0.3089229464530945, "rewards/simpleverify_reward/mean": 0.41015625, "rewards/simpleverify_reward/std": 0.49234291911125183, "step": 440, "tools/generated_tokens": 2197.310546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.513671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4765625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1940.0, "completions/mean_length": 1528.87109375, "completions/mean_terminated_length": 1056.2313232421875, "completions/min_length": 86.0, "completions/min_terminated_length": 86.0, "epoch": 0.3129879347054649, "frac_reward_zero_std": 0.3125, "grad_norm": 2.1062240600585938, "learning_rate": 1.2e-07, "loss": 0.0725, "num_tokens": 1004355920.0, "reward": 0.30859375, "reward_std": 0.28357410430908203, "rewards/simpleverify_reward/mean": 0.30859375, "rewards/simpleverify_reward/std": 0.4623647928237915, "step": 441, "tools/generated_tokens": 2843.84765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.72265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.470703125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1981.0, "completions/mean_length": 1520.67578125, "completions/mean_terminated_length": 1058.830322265625, "completions/min_length": 104.0, "completions/min_terminated_length": 104.0, "epoch": 0.31369765791341375, "frac_reward_zero_std": 0.28125, "grad_norm": 2.815082550048828, "learning_rate": 1.1799999999999998e-07, "loss": 0.068, "num_tokens": 1006607922.0, "reward": 0.2421875, "reward_std": 0.28135883808135986, "rewards/simpleverify_reward/mean": 0.2421875, "rewards/simpleverify_reward/std": 0.42882615327835083, "step": 442, "tools/generated_tokens": 2700.70703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.646484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.267578125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1982.0, "completions/mean_length": 1322.365234375, "completions/mean_terminated_length": 1057.61328125, "completions/min_length": 240.0, "completions/min_terminated_length": 240.0, "epoch": 0.3144073811213627, "frac_reward_zero_std": 0.15625, "grad_norm": 1.8645985126495361, "learning_rate": 1.16e-07, "loss": 0.0631, "num_tokens": 1008589541.0, "reward": 0.49609375, "reward_std": 0.3681023120880127, "rewards/simpleverify_reward/mean": 0.49609375, "rewards/simpleverify_reward/std": 0.5004737377166748, "step": 443, "tools/generated_tokens": 2295.64453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5234375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.28515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1919.0, "completions/mean_length": 1329.791015625, "completions/mean_terminated_length": 1044.1502685546875, "completions/min_length": 1.0, "completions/min_terminated_length": 1.0, "epoch": 0.31511710432931156, "frac_reward_zero_std": 0.1875, "grad_norm": 2.078202962875366, "learning_rate": 1.14e-07, "loss": 0.0565, "num_tokens": 1010490214.0, "reward": 0.48828125, "reward_std": 0.31245148181915283, "rewards/simpleverify_reward/mean": 0.48828125, "rewards/simpleverify_reward/std": 0.5003514885902405, "step": 444, "tools/generated_tokens": 2219.455078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4296875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2019.0, "completions/mean_length": 1533.87890625, "completions/mean_terminated_length": 1150.986328125, "completions/min_length": 223.0, "completions/min_terminated_length": 223.0, "epoch": 0.3158268275372605, "frac_reward_zero_std": 0.25, "grad_norm": 2.4454658031463623, "learning_rate": 1.12e-07, "loss": 0.0323, "num_tokens": 1012820419.0, "reward": 0.337890625, "reward_std": 0.30881643295288086, "rewards/simpleverify_reward/mean": 0.337890625, "rewards/simpleverify_reward/std": 0.4734536409378052, "step": 445, "tools/generated_tokens": 2817.01953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.318359375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1960.0, "completions/mean_length": 1347.779296875, "completions/mean_terminated_length": 1021.54443359375, "completions/min_length": 69.0, "completions/min_terminated_length": 69.0, "epoch": 0.31653655074520937, "frac_reward_zero_std": 0.34375, "grad_norm": 2.1163806915283203, "learning_rate": 1.0999999999999999e-07, "loss": 0.0538, "num_tokens": 1014873562.0, "reward": 0.447265625, "reward_std": 0.2908296585083008, "rewards/simpleverify_reward/mean": 0.447265625, "rewards/simpleverify_reward/std": 0.4976975917816162, "step": 446, "tools/generated_tokens": 2397.4296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.568359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.314453125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1914.0, "completions/mean_length": 1309.34765625, "completions/mean_terminated_length": 970.5355834960938, "completions/min_length": 58.0, "completions/min_terminated_length": 58.0, "epoch": 0.31724627395315824, "frac_reward_zero_std": 0.3125, "grad_norm": 2.3217012882232666, "learning_rate": 1.0799999999999999e-07, "loss": 0.0544, "num_tokens": 1016762761.0, "reward": 0.4765625, "reward_std": 0.26956161856651306, "rewards/simpleverify_reward/mean": 0.4765625, "rewards/simpleverify_reward/std": 0.49993884563446045, "step": 447, "tools/generated_tokens": 2185.73828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.478515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1947.0, "completions/mean_length": 1477.126953125, "completions/mean_terminated_length": 1099.1883544921875, "completions/min_length": 44.0, "completions/min_terminated_length": 44.0, "epoch": 0.3179559971611072, "frac_reward_zero_std": 0.3125, "grad_norm": 2.2992076873779297, "learning_rate": 1.06e-07, "loss": 0.0563, "num_tokens": 1019033948.0, "reward": 0.37109375, "reward_std": 0.2795700132846832, "rewards/simpleverify_reward/mean": 0.37109375, "rewards/simpleverify_reward/std": 0.4835699498653412, "step": 448, "tools/generated_tokens": 2715.275390625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.66796875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2003.0, "completions/mean_length": 1353.435546875, "completions/mean_terminated_length": 1015.465087890625, "completions/min_length": 93.0, "completions/min_terminated_length": 93.0, "epoch": 0.31866572036905605, "frac_reward_zero_std": 0.21875, "grad_norm": 3.643096446990967, "learning_rate": 1.0399999999999999e-07, "loss": 0.0462, "num_tokens": 1021086521.0, "reward": 0.427734375, "reward_std": 0.33298587799072266, "rewards/simpleverify_reward/mean": 0.427734375, "rewards/simpleverify_reward/std": 0.4952339828014374, "step": 449, "tools/generated_tokens": 2399.63671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.5, "completions/max_length": 2048.0, "completions/max_terminated_length": 1988.0, "completions/mean_length": 1585.484375, "completions/mean_terminated_length": 1125.40625, "completions/min_length": 1.0, "completions/min_terminated_length": 1.0, "epoch": 0.319375443577005, "frac_reward_zero_std": 0.25, "grad_norm": 3.114999294281006, "learning_rate": 1.0199999999999999e-07, "loss": 0.0696, "num_tokens": 1023454176.0, "reward": 0.294921875, "reward_std": 0.29965174198150635, "rewards/simpleverify_reward/mean": 0.294921875, "rewards/simpleverify_reward/std": 0.4564536213874817, "step": 450, "tools/generated_tokens": 2859.37109375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.701171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.322265625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1992.0, "completions/mean_length": 1374.041015625, "completions/mean_terminated_length": 1053.570556640625, "completions/min_length": 84.0, "completions/min_terminated_length": 84.0, "epoch": 0.32008516678495386, "frac_reward_zero_std": 0.15625, "grad_norm": 2.2116236686706543, "learning_rate": 1e-07, "loss": 0.0621, "num_tokens": 1025558699.0, "reward": 0.38671875, "reward_std": 0.34369954466819763, "rewards/simpleverify_reward/mean": 0.38671875, "rewards/simpleverify_reward/std": 0.48747459053993225, "step": 451, "tools/generated_tokens": 2489.6328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.60546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.30078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1921.0, "completions/mean_length": 1378.490234375, "completions/mean_terminated_length": 1091.2430419921875, "completions/min_length": 34.0, "completions/min_terminated_length": 34.0, "epoch": 0.3207948899929028, "frac_reward_zero_std": 0.28125, "grad_norm": 1.704269289970398, "learning_rate": 9.8e-08, "loss": 0.0636, "num_tokens": 1027527908.0, "reward": 0.470703125, "reward_std": 0.27617377042770386, "rewards/simpleverify_reward/mean": 0.470703125, "rewards/simpleverify_reward/std": 0.49962911009788513, "step": 452, "tools/generated_tokens": 2314.23046875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.509765625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.294921875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1989.0, "completions/mean_length": 1329.376953125, "completions/mean_terminated_length": 1031.2464599609375, "completions/min_length": 103.0, "completions/min_terminated_length": 103.0, "epoch": 0.32150461320085166, "frac_reward_zero_std": 0.375, "grad_norm": 2.728440046310425, "learning_rate": 9.6e-08, "loss": 0.0605, "num_tokens": 1029552815.0, "reward": 0.412109375, "reward_std": 0.2412969172000885, "rewards/simpleverify_reward/mean": 0.412109375, "rewards/simpleverify_reward/std": 0.49269601702690125, "step": 453, "tools/generated_tokens": 2389.30859375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.572265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.373046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1959.0, "completions/mean_length": 1345.68359375, "completions/mean_terminated_length": 928.7694702148438, "completions/min_length": 106.0, "completions/min_terminated_length": 106.0, "epoch": 0.3222143364088006, "frac_reward_zero_std": 0.3125, "grad_norm": 1.757079005241394, "learning_rate": 9.4e-08, "loss": 0.0676, "num_tokens": 1031640860.0, "reward": 0.44140625, "reward_std": 0.27982833981513977, "rewards/simpleverify_reward/mean": 0.44140625, "rewards/simpleverify_reward/std": 0.4970405399799347, "step": 454, "tools/generated_tokens": 2439.185546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.59375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.21875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2047.0, "completions/mean_length": 1248.197265625, "completions/mean_terminated_length": 1025.25244140625, "completions/min_length": 106.0, "completions/min_terminated_length": 106.0, "epoch": 0.32292405961674947, "frac_reward_zero_std": 0.40625, "grad_norm": 1.1127564907073975, "learning_rate": 9.199999999999999e-08, "loss": 0.0383, "num_tokens": 1033416932.0, "reward": 0.552734375, "reward_std": 0.2269904911518097, "rewards/simpleverify_reward/mean": 0.552734375, "rewards/simpleverify_reward/std": 0.4976975917816162, "step": 455, "tools/generated_tokens": 2039.46875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.427734375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1911.0, "completions/mean_length": 1444.361328125, "completions/mean_terminated_length": 1117.0872802734375, "completions/min_length": 103.0, "completions/min_terminated_length": 103.0, "epoch": 0.32363378282469835, "frac_reward_zero_std": 0.125, "grad_norm": 2.6580474376678467, "learning_rate": 9e-08, "loss": 0.0768, "num_tokens": 1035678544.0, "reward": 0.392578125, "reward_std": 0.3405269682407379, "rewards/simpleverify_reward/mean": 0.392578125, "rewards/simpleverify_reward/std": 0.4888018071651459, "step": 456, "tools/generated_tokens": 2667.94921875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.666015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.357421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2008.0, "completions/mean_length": 1451.04296875, "completions/mean_terminated_length": 1118.9969482421875, "completions/min_length": 166.0, "completions/min_terminated_length": 166.0, "epoch": 0.3243435060326473, "frac_reward_zero_std": 0.28125, "grad_norm": 2.6185789108276367, "learning_rate": 8.8e-08, "loss": 0.0568, "num_tokens": 1037834773.0, "reward": 0.369140625, "reward_std": 0.3223860263824463, "rewards/simpleverify_reward/mean": 0.369140625, "rewards/simpleverify_reward/std": 0.4830440282821655, "step": 457, "tools/generated_tokens": 2561.521484375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.35546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2016.0, "completions/mean_length": 1382.6171875, "completions/mean_terminated_length": 1019.066650390625, "completions/min_length": 120.0, "completions/min_terminated_length": 120.0, "epoch": 0.32505322924059615, "frac_reward_zero_std": 0.28125, "grad_norm": 1.518701434135437, "learning_rate": 8.599999999999999e-08, "loss": 0.0681, "num_tokens": 1039826417.0, "reward": 0.41015625, "reward_std": 0.30442410707473755, "rewards/simpleverify_reward/mean": 0.41015625, "rewards/simpleverify_reward/std": 0.49234291911125183, "step": 458, "tools/generated_tokens": 2347.017578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.52734375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.310546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1950.0, "completions/mean_length": 1385.53515625, "completions/mean_terminated_length": 1088.0311279296875, "completions/min_length": 171.0, "completions/min_terminated_length": 171.0, "epoch": 0.3257629524485451, "frac_reward_zero_std": 0.21875, "grad_norm": 9.823101043701172, "learning_rate": 8.4e-08, "loss": 0.0707, "num_tokens": 1041880759.0, "reward": 0.42578125, "reward_std": 0.3140801787376404, "rewards/simpleverify_reward/mean": 0.42578125, "rewards/simpleverify_reward/std": 0.4949444830417633, "step": 459, "tools/generated_tokens": 2396.6015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.568359375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.30078125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1956.0, "completions/mean_length": 1390.625, "completions/mean_terminated_length": 1108.0977783203125, "completions/min_length": 106.0, "completions/min_terminated_length": 106.0, "epoch": 0.32647267565649396, "frac_reward_zero_std": 0.1875, "grad_norm": 2.601546049118042, "learning_rate": 8.2e-08, "loss": 0.0256, "num_tokens": 1043901807.0, "reward": 0.33984375, "reward_std": 0.30809134244918823, "rewards/simpleverify_reward/mean": 0.33984375, "rewards/simpleverify_reward/std": 0.4741191864013672, "step": 460, "tools/generated_tokens": 2398.85546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.544921875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1885.0, "completions/mean_length": 1436.4140625, "completions/mean_terminated_length": 1060.746826171875, "completions/min_length": 263.0, "completions/min_terminated_length": 263.0, "epoch": 0.3271823988644429, "frac_reward_zero_std": 0.34375, "grad_norm": 1.48546302318573, "learning_rate": 8e-08, "loss": 0.0342, "num_tokens": 1046021803.0, "reward": 0.388671875, "reward_std": 0.2336704432964325, "rewards/simpleverify_reward/mean": 0.388671875, "rewards/simpleverify_reward/std": 0.4879252314567566, "step": 461, "tools/generated_tokens": 2530.27734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.59375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.298828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1962.0, "completions/mean_length": 1365.96875, "completions/mean_terminated_length": 1075.598876953125, "completions/min_length": 205.0, "completions/min_terminated_length": 205.0, "epoch": 0.32789212207239177, "frac_reward_zero_std": 0.3125, "grad_norm": 2.037592887878418, "learning_rate": 7.8e-08, "loss": 0.0524, "num_tokens": 1048033480.0, "reward": 0.46484375, "reward_std": 0.26226192712783813, "rewards/simpleverify_reward/mean": 0.46484375, "rewards/simpleverify_reward/std": 0.49925029277801514, "step": 462, "tools/generated_tokens": 2377.11328125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.361328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1965.0, "completions/mean_length": 1406.24609375, "completions/mean_terminated_length": 1044.12841796875, "completions/min_length": 119.0, "completions/min_terminated_length": 119.0, "epoch": 0.32860184528034064, "frac_reward_zero_std": 0.34375, "grad_norm": 1.6462810039520264, "learning_rate": 7.599999999999999e-08, "loss": 0.0577, "num_tokens": 1050183254.0, "reward": 0.27734375, "reward_std": 0.2607883810997009, "rewards/simpleverify_reward/mean": 0.27734375, "rewards/simpleverify_reward/std": 0.4481254518032074, "step": 463, "tools/generated_tokens": 2543.107421875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.326171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2012.0, "completions/mean_length": 1309.408203125, "completions/mean_terminated_length": 953.0319213867188, "completions/min_length": 151.0, "completions/min_terminated_length": 151.0, "epoch": 0.3293115684882896, "frac_reward_zero_std": 0.46875, "grad_norm": 1.4736710786819458, "learning_rate": 7.399999999999999e-08, "loss": 0.067, "num_tokens": 1052174337.0, "reward": 0.478515625, "reward_std": 0.20495682954788208, "rewards/simpleverify_reward/mean": 0.478515625, "rewards/simpleverify_reward/std": 0.5000267624855042, "step": 464, "tools/generated_tokens": 2298.70703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.54296875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.337890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1999.0, "completions/mean_length": 1304.837890625, "completions/mean_terminated_length": 927.6990966796875, "completions/min_length": 143.0, "completions/min_terminated_length": 143.0, "epoch": 0.33002129169623845, "frac_reward_zero_std": 0.46875, "grad_norm": 1.9416823387145996, "learning_rate": 7.2e-08, "loss": 0.0648, "num_tokens": 1054156453.0, "reward": 0.39453125, "reward_std": 0.23405250906944275, "rewards/simpleverify_reward/mean": 0.39453125, "rewards/simpleverify_reward/std": 0.4892277717590332, "step": 465, "tools/generated_tokens": 2285.81640625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.533203125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1940.0, "completions/mean_length": 1443.484375, "completions/mean_terminated_length": 1017.7933349609375, "completions/min_length": 110.0, "completions/min_terminated_length": 110.0, "epoch": 0.3307310149041874, "frac_reward_zero_std": 0.34375, "grad_norm": 1.4031037092208862, "learning_rate": 7e-08, "loss": 0.0741, "num_tokens": 1056336300.0, "reward": 0.36328125, "reward_std": 0.2962000370025635, "rewards/simpleverify_reward/mean": 0.36328125, "rewards/simpleverify_reward/std": 0.4814152419567108, "step": 466, "tools/generated_tokens": 2589.744140625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.62890625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.376953125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1938.0, "completions/mean_length": 1445.728515625, "completions/mean_terminated_length": 1081.3448486328125, "completions/min_length": 79.0, "completions/min_terminated_length": 79.0, "epoch": 0.33144073811213626, "frac_reward_zero_std": 0.21875, "grad_norm": 3.18921160697937, "learning_rate": 6.8e-08, "loss": 0.0975, "num_tokens": 1058548962.0, "reward": 0.462890625, "reward_std": 0.34283381700515747, "rewards/simpleverify_reward/mean": 0.462890625, "rewards/simpleverify_reward/std": 0.4991086423397064, "step": 467, "tools/generated_tokens": 2653.693359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.654296875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3984375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1998.0, "completions/mean_length": 1455.5859375, "completions/mean_terminated_length": 1063.207763671875, "completions/min_length": 69.0, "completions/min_terminated_length": 69.0, "epoch": 0.3321504613200852, "frac_reward_zero_std": 0.25, "grad_norm": 2.9189767837524414, "learning_rate": 6.6e-08, "loss": 0.0631, "num_tokens": 1060766665.0, "reward": 0.4140625, "reward_std": 0.30618906021118164, "rewards/simpleverify_reward/mean": 0.4140625, "rewards/simpleverify_reward/std": 0.49304109811782837, "step": 468, "tools/generated_tokens": 2629.185546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.63671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.30859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1999.0, "completions/mean_length": 1322.4453125, "completions/mean_terminated_length": 999.6044921875, "completions/min_length": 78.0, "completions/min_terminated_length": 78.0, "epoch": 0.33286018452803406, "frac_reward_zero_std": 0.3125, "grad_norm": 2.492673873901367, "learning_rate": 6.4e-08, "loss": 0.055, "num_tokens": 1062709856.0, "reward": 0.4453125, "reward_std": 0.30089086294174194, "rewards/simpleverify_reward/mean": 0.4453125, "rewards/simpleverify_reward/std": 0.49748632311820984, "step": 469, "tools/generated_tokens": 2245.95703125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.501953125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1954.0, "completions/mean_length": 1460.3359375, "completions/mean_terminated_length": 1031.5, "completions/min_length": 108.0, "completions/min_terminated_length": 108.0, "epoch": 0.33356990773598294, "frac_reward_zero_std": 0.4375, "grad_norm": 2.537611722946167, "learning_rate": 6.2e-08, "loss": 0.0434, "num_tokens": 1065078457.0, "reward": 0.31640625, "reward_std": 0.2361874282360077, "rewards/simpleverify_reward/mean": 0.31640625, "rewards/simpleverify_reward/std": 0.46552830934524536, "step": 470, "tools/generated_tokens": 2827.46875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.7421875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1982.0, "completions/mean_length": 1391.2734375, "completions/mean_terminated_length": 1010.2098999023438, "completions/min_length": 109.0, "completions/min_terminated_length": 109.0, "epoch": 0.33427963094393187, "frac_reward_zero_std": 0.15625, "grad_norm": 2.2226502895355225, "learning_rate": 6e-08, "loss": 0.0688, "num_tokens": 1067242233.0, "reward": 0.419921875, "reward_std": 0.3496588468551636, "rewards/simpleverify_reward/mean": 0.419921875, "rewards/simpleverify_reward/std": 0.4940285086631775, "step": 471, "tools/generated_tokens": 2541.412109375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.623046875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1995.0, "completions/mean_length": 1464.798828125, "completions/mean_terminated_length": 1103.24365234375, "completions/min_length": 37.0, "completions/min_terminated_length": 37.0, "epoch": 0.33498935415188075, "frac_reward_zero_std": 0.28125, "grad_norm": 1.675835132598877, "learning_rate": 5.8e-08, "loss": 0.008, "num_tokens": 1069390040.0, "reward": 0.345703125, "reward_std": 0.28101515769958496, "rewards/simpleverify_reward/mean": 0.345703125, "rewards/simpleverify_reward/std": 0.4760620892047882, "step": 472, "tools/generated_tokens": 2567.822265625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.353515625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1982.0, "completions/mean_length": 1444.052734375, "completions/mean_terminated_length": 1115.265869140625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.3356990773598297, "frac_reward_zero_std": 0.21875, "grad_norm": 1.60427725315094, "learning_rate": 5.6e-08, "loss": 0.0439, "num_tokens": 1071485749.0, "reward": 0.408203125, "reward_std": 0.31041550636291504, "rewards/simpleverify_reward/mean": 0.408203125, "rewards/simpleverify_reward/std": 0.49198177456855774, "step": 473, "tools/generated_tokens": 2506.51953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.576171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.275390625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1947.0, "completions/mean_length": 1349.767578125, "completions/mean_terminated_length": 1085.1185302734375, "completions/min_length": 2.0, "completions/min_terminated_length": 2.0, "epoch": 0.33640880056777855, "frac_reward_zero_std": 0.3125, "grad_norm": 1.6276025772094727, "learning_rate": 5.3999999999999994e-08, "loss": 0.0876, "num_tokens": 1073398880.0, "reward": 0.49609375, "reward_std": 0.28104907274246216, "rewards/simpleverify_reward/mean": 0.49609375, "rewards/simpleverify_reward/std": 0.5004737377166748, "step": 474, "tools/generated_tokens": 2263.751953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.49609375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1989.0, "completions/mean_length": 1387.56640625, "completions/mean_terminated_length": 978.5253295898438, "completions/min_length": 125.0, "completions/min_terminated_length": 125.0, "epoch": 0.3371185237757275, "frac_reward_zero_std": 0.375, "grad_norm": 2.3629963397979736, "learning_rate": 5.1999999999999996e-08, "loss": 0.056, "num_tokens": 1075574444.0, "reward": 0.388671875, "reward_std": 0.2854521870613098, "rewards/simpleverify_reward/mean": 0.388671875, "rewards/simpleverify_reward/std": 0.4879252314567566, "step": 475, "tools/generated_tokens": 2567.6953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.646484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.224609375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1961.0, "completions/mean_length": 1248.306640625, "completions/mean_terminated_length": 1018.4105224609375, "completions/min_length": 72.0, "completions/min_terminated_length": 72.0, "epoch": 0.33782824698367636, "frac_reward_zero_std": 0.4375, "grad_norm": 1.4819141626358032, "learning_rate": 5e-08, "loss": 0.0489, "num_tokens": 1077251614.0, "reward": 0.45703125, "reward_std": 0.2255207896232605, "rewards/simpleverify_reward/mean": 0.45703125, "rewards/simpleverify_reward/std": 0.49863746762275696, "step": 476, "tools/generated_tokens": 1919.029296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.365234375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.423828125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1951.0, "completions/mean_length": 1449.94140625, "completions/mean_terminated_length": 1011.23388671875, "completions/min_length": 260.0, "completions/min_terminated_length": 260.0, "epoch": 0.3385379701916253, "frac_reward_zero_std": 0.28125, "grad_norm": 2.5130934715270996, "learning_rate": 4.8e-08, "loss": 0.0809, "num_tokens": 1079417547.0, "reward": 0.396484375, "reward_std": 0.29014095664024353, "rewards/simpleverify_reward/mean": 0.396484375, "rewards/simpleverify_reward/std": 0.4896455705165863, "step": 477, "tools/generated_tokens": 2549.509765625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.326171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1980.0, "completions/mean_length": 1354.384765625, "completions/mean_terminated_length": 1022.8580322265625, "completions/min_length": 109.0, "completions/min_terminated_length": 109.0, "epoch": 0.33924769339957417, "frac_reward_zero_std": 0.375, "grad_norm": 1.81350839138031, "learning_rate": 4.5999999999999995e-08, "loss": 0.0295, "num_tokens": 1081436154.0, "reward": 0.34765625, "reward_std": 0.26024550199508667, "rewards/simpleverify_reward/mean": 0.34765625, "rewards/simpleverify_reward/std": 0.47669193148612976, "step": 478, "tools/generated_tokens": 2369.263671875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.55078125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.390625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1983.0, "completions/mean_length": 1444.498046875, "completions/mean_terminated_length": 1058.6121826171875, "completions/min_length": 38.0, "completions/min_terminated_length": 38.0, "epoch": 0.33995741660752304, "frac_reward_zero_std": 0.3125, "grad_norm": 1.2183293104171753, "learning_rate": 4.4e-08, "loss": 0.0606, "num_tokens": 1083585641.0, "reward": 0.30859375, "reward_std": 0.2647823393344879, "rewards/simpleverify_reward/mean": 0.30859375, "rewards/simpleverify_reward/std": 0.4623647928237915, "step": 479, "tools/generated_tokens": 2553.189453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.60546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.35546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1948.0, "completions/mean_length": 1407.927734375, "completions/mean_terminated_length": 1055.0, "completions/min_length": 176.0, "completions/min_terminated_length": 176.0, "epoch": 0.340667139815472, "frac_reward_zero_std": 0.375, "grad_norm": 2.0719683170318604, "learning_rate": 4.2e-08, "loss": 0.0436, "num_tokens": 1085618251.0, "reward": 0.474609375, "reward_std": 0.2424173355102539, "rewards/simpleverify_reward/mean": 0.474609375, "rewards/simpleverify_reward/std": 0.4998432695865631, "step": 480, "tools/generated_tokens": 2418.197265625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.55078125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.357421875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2021.0, "completions/mean_length": 1416.763671875, "completions/mean_terminated_length": 1066.924072265625, "completions/min_length": 43.0, "completions/min_terminated_length": 43.0, "epoch": 0.34137686302342085, "frac_reward_zero_std": 0.1875, "grad_norm": 4.262824058532715, "learning_rate": 4e-08, "loss": 0.0736, "num_tokens": 1087744198.0, "reward": 0.373046875, "reward_std": 0.3384213447570801, "rewards/simpleverify_reward/mean": 0.373046875, "rewards/simpleverify_reward/std": 0.48408737778663635, "step": 481, "tools/generated_tokens": 2527.15625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.607421875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.369140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2003.0, "completions/mean_length": 1417.884765625, "completions/mean_terminated_length": 1049.1795654296875, "completions/min_length": 46.0, "completions/min_terminated_length": 46.0, "epoch": 0.3420865862313698, "frac_reward_zero_std": 0.21875, "grad_norm": 2.3229451179504395, "learning_rate": 3.7999999999999996e-08, "loss": 0.0649, "num_tokens": 1089950364.0, "reward": 0.298828125, "reward_std": 0.33337730169296265, "rewards/simpleverify_reward/mean": 0.298828125, "rewards/simpleverify_reward/std": 0.45819199085235596, "step": 482, "tools/generated_tokens": 2628.748046875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.65625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.2890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1937.0, "completions/mean_length": 1346.09375, "completions/mean_terminated_length": 1061.71435546875, "completions/min_length": 135.0, "completions/min_terminated_length": 135.0, "epoch": 0.34279630943931866, "frac_reward_zero_std": 0.15625, "grad_norm": 2.096663475036621, "learning_rate": 3.6e-08, "loss": 0.0563, "num_tokens": 1091990177.0, "reward": 0.3515625, "reward_std": 0.33231645822525024, "rewards/simpleverify_reward/mean": 0.3515625, "rewards/simpleverify_reward/std": 0.4779251217842102, "step": 483, "tools/generated_tokens": 2413.322265625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.435546875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1988.0, "completions/mean_length": 1463.56640625, "completions/mean_terminated_length": 1013.6332397460938, "completions/min_length": 129.0, "completions/min_terminated_length": 129.0, "epoch": 0.3435060326472676, "frac_reward_zero_std": 0.46875, "grad_norm": 0.9510486125946045, "learning_rate": 3.4e-08, "loss": 0.056, "num_tokens": 1094182993.0, "reward": 0.30859375, "reward_std": 0.21385458111763, "rewards/simpleverify_reward/mean": 0.30859375, "rewards/simpleverify_reward/std": 0.4623647928237915, "step": 484, "tools/generated_tokens": 2608.953125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.369140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1935.0, "completions/mean_length": 1339.24609375, "completions/mean_terminated_length": 924.5263061523438, "completions/min_length": 135.0, "completions/min_terminated_length": 135.0, "epoch": 0.34421575585521647, "frac_reward_zero_std": 0.125, "grad_norm": 2.2297990322113037, "learning_rate": 3.2e-08, "loss": 0.0743, "num_tokens": 1096219662.0, "reward": 0.353515625, "reward_std": 0.3416631519794464, "rewards/simpleverify_reward/mean": 0.353515625, "rewards/simpleverify_reward/std": 0.47852855920791626, "step": 485, "tools/generated_tokens": 2376.80078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.3671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1980.0, "completions/mean_length": 1406.296875, "completions/mean_terminated_length": 1033.95068359375, "completions/min_length": 152.0, "completions/min_terminated_length": 152.0, "epoch": 0.34492547906316534, "frac_reward_zero_std": 0.28125, "grad_norm": 2.028433322906494, "learning_rate": 3e-08, "loss": 0.069, "num_tokens": 1098278485.0, "reward": 0.404296875, "reward_std": 0.266385555267334, "rewards/simpleverify_reward/mean": 0.404296875, "rewards/simpleverify_reward/std": 0.4912354052066803, "step": 486, "tools/generated_tokens": 2433.318359375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.560546875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.33203125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2026.0, "completions/mean_length": 1368.123046875, "completions/mean_terminated_length": 1031.0614013671875, "completions/min_length": 30.0, "completions/min_terminated_length": 30.0, "epoch": 0.3456352022711143, "frac_reward_zero_std": 0.25, "grad_norm": 2.0322349071502686, "learning_rate": 2.8e-08, "loss": 0.0741, "num_tokens": 1100361780.0, "reward": 0.451171875, "reward_std": 0.3090813457965851, "rewards/simpleverify_reward/mean": 0.451171875, "rewards/simpleverify_reward/std": 0.498096764087677, "step": 487, "tools/generated_tokens": 2452.05078125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.5859375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.369140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1927.0, "completions/mean_length": 1518.92578125, "completions/mean_terminated_length": 1212.640869140625, "completions/min_length": 23.0, "completions/min_terminated_length": 23.0, "epoch": 0.34634492547906315, "frac_reward_zero_std": 0.1875, "grad_norm": 2.106015205383301, "learning_rate": 2.5999999999999998e-08, "loss": 0.063, "num_tokens": 1102646885.0, "reward": 0.39453125, "reward_std": 0.3605169951915741, "rewards/simpleverify_reward/mean": 0.39453125, "rewards/simpleverify_reward/std": 0.4892277717590332, "step": 488, "tools/generated_tokens": 2765.73828125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.673828125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.388671875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2001.0, "completions/mean_length": 1442.794921875, "completions/mean_terminated_length": 1058.0159912109375, "completions/min_length": 50.0, "completions/min_terminated_length": 50.0, "epoch": 0.3470546486870121, "frac_reward_zero_std": 0.3125, "grad_norm": 2.0192654132843018, "learning_rate": 2.4e-08, "loss": 0.0614, "num_tokens": 1104834331.0, "reward": 0.375, "reward_std": 0.28136348724365234, "rewards/simpleverify_reward/mean": 0.375, "rewards/simpleverify_reward/std": 0.4845963716506958, "step": 489, "tools/generated_tokens": 2608.64453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.6328125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.001953125 }, { "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": 0.427734375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1959.0, "completions/mean_length": 1471.353515625, "completions/mean_terminated_length": 1043.498291015625, "completions/min_length": 152.0, "completions/min_terminated_length": 152.0, "epoch": 0.34776437189496096, "frac_reward_zero_std": 0.34375, "grad_norm": 1.9250950813293457, "learning_rate": 2.2e-08, "loss": 0.0586, "num_tokens": 1107133250.0, "reward": 0.28515625, "reward_std": 0.25588494539260864, "rewards/simpleverify_reward/mean": 0.28515625, "rewards/simpleverify_reward/std": 0.45193037390708923, "step": 490, "tools/generated_tokens": 2710.267578125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.671875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.337890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1957.0, "completions/mean_length": 1327.521484375, "completions/mean_terminated_length": 961.07080078125, "completions/min_length": 174.0, "completions/min_terminated_length": 174.0, "epoch": 0.3484740951029099, "frac_reward_zero_std": 0.21875, "grad_norm": 2.584733009338379, "learning_rate": 2e-08, "loss": 0.0598, "num_tokens": 1109119782.0, "reward": 0.46484375, "reward_std": 0.3010560870170593, "rewards/simpleverify_reward/mean": 0.46484375, "rewards/simpleverify_reward/std": 0.49925029277801514, "step": 491, "tools/generated_tokens": 2315.78125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.541015625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.36328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1979.0, "completions/mean_length": 1351.517578125, "completions/mean_terminated_length": 957.2024536132812, "completions/min_length": 96.0, "completions/min_terminated_length": 96.0, "epoch": 0.34918381831085876, "frac_reward_zero_std": 0.1875, "grad_norm": 1.804661512374878, "learning_rate": 1.8e-08, "loss": 0.0719, "num_tokens": 1111174244.0, "reward": 0.42578125, "reward_std": 0.33554884791374207, "rewards/simpleverify_reward/mean": 0.42578125, "rewards/simpleverify_reward/std": 0.4949444830417633, "step": 492, "tools/generated_tokens": 2406.283203125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.576171875, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.328125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1984.0, "completions/mean_length": 1352.0, "completions/mean_terminated_length": 1013.1162719726562, "completions/min_length": 180.0, "completions/min_terminated_length": 180.0, "epoch": 0.34989354151880764, "frac_reward_zero_std": 0.28125, "grad_norm": 1.472298264503479, "learning_rate": 1.6e-08, "loss": 0.0412, "num_tokens": 1113184464.0, "reward": 0.498046875, "reward_std": 0.29150667786598206, "rewards/simpleverify_reward/mean": 0.498046875, "rewards/simpleverify_reward/std": 0.5004851818084717, "step": 493, "tools/generated_tokens": 2363.66015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.548828125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.4140625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2017.0, "completions/mean_length": 1487.97265625, "completions/mean_terminated_length": 1093.9200439453125, "completions/min_length": 111.0, "completions/min_terminated_length": 111.0, "epoch": 0.35060326472675657, "frac_reward_zero_std": 0.1875, "grad_norm": 8.202691078186035, "learning_rate": 1.4e-08, "loss": 0.0723, "num_tokens": 1115534587.0, "reward": 0.369140625, "reward_std": 0.33920201659202576, "rewards/simpleverify_reward/mean": 0.369140625, "rewards/simpleverify_reward/std": 0.4830440282821655, "step": 494, "tools/generated_tokens": 2806.810546875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.72265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.326171875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1947.0, "completions/mean_length": 1431.1875, "completions/mean_terminated_length": 1132.6754150390625, "completions/min_length": 133.0, "completions/min_terminated_length": 133.0, "epoch": 0.35131298793470545, "frac_reward_zero_std": 0.375, "grad_norm": 1.478010892868042, "learning_rate": 1.2e-08, "loss": 0.0781, "num_tokens": 1117657335.0, "reward": 0.46875, "reward_std": 0.27191591262817383, "rewards/simpleverify_reward/mean": 0.46875, "rewards/simpleverify_reward/std": 0.4995105266571045, "step": 495, "tools/generated_tokens": 2524.5859375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.595703125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.251953125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2031.0, "completions/mean_length": 1300.078125, "completions/mean_terminated_length": 1049.0966796875, "completions/min_length": 73.0, "completions/min_terminated_length": 73.0, "epoch": 0.3520227111426544, "frac_reward_zero_std": 0.28125, "grad_norm": 2.052149534225464, "learning_rate": 1e-08, "loss": 0.0817, "num_tokens": 1119586135.0, "reward": 0.521484375, "reward_std": 0.27696457505226135, "rewards/simpleverify_reward/mean": 0.521484375, "rewards/simpleverify_reward/std": 0.5000267624855042, "step": 496, "tools/generated_tokens": 2246.04296875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.517578125, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.337890625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2002.0, "completions/mean_length": 1409.349609375, "completions/mean_terminated_length": 1083.4306640625, "completions/min_length": 79.0, "completions/min_terminated_length": 79.0, "epoch": 0.35273243435060325, "frac_reward_zero_std": 0.21875, "grad_norm": 3.3236312866210938, "learning_rate": 8e-09, "loss": 0.1084, "num_tokens": 1121652541.0, "reward": 0.400390625, "reward_std": 0.3154503107070923, "rewards/simpleverify_reward/mean": 0.400390625, "rewards/simpleverify_reward/std": 0.4904567301273346, "step": 497, "tools/generated_tokens": 2454.716796875, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.572265625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.255859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 2034.0, "completions/mean_length": 1289.435546875, "completions/mean_terminated_length": 1029.1207275390625, "completions/min_length": 45.0, "completions/min_terminated_length": 45.0, "epoch": 0.3534421575585522, "frac_reward_zero_std": 0.25, "grad_norm": 3.2848219871520996, "learning_rate": 6e-09, "loss": 0.0962, "num_tokens": 1123530394.0, "reward": 0.427734375, "reward_std": 0.3406909108161926, "rewards/simpleverify_reward/mean": 0.427734375, "rewards/simpleverify_reward/std": 0.4952339828014374, "step": 498, "tools/generated_tokens": 2182.166015625, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.484375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.291015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 2014.0, "completions/mean_length": 1375.068359375, "completions/mean_terminated_length": 1099.942138671875, "completions/min_length": 188.0, "completions/min_terminated_length": 188.0, "epoch": 0.35415188076650106, "frac_reward_zero_std": 0.21875, "grad_norm": 1.6319494247436523, "learning_rate": 4e-09, "loss": 0.0887, "num_tokens": 1125625949.0, "reward": 0.404296875, "reward_std": 0.3444260358810425, "rewards/simpleverify_reward/mean": 0.404296875, "rewards/simpleverify_reward/std": 0.4912354052066803, "step": 499, "tools/generated_tokens": 2463.302734375, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.58984375, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "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": 0.380859375, "completions/max_length": 2048.0, "completions/max_terminated_length": 1974.0, "completions/mean_length": 1414.404296875, "completions/mean_terminated_length": 1026.4605712890625, "completions/min_length": 188.0, "completions/min_terminated_length": 188.0, "epoch": 0.35486160397445, "frac_reward_zero_std": 0.46875, "grad_norm": 1.0372529029846191, "learning_rate": 2e-09, "loss": 0.0322, "num_tokens": 1127757057.0, "reward": 0.34765625, "reward_std": 0.2021893560886383, "rewards/simpleverify_reward/mean": 0.34765625, "rewards/simpleverify_reward/std": 0.47669193148612976, "step": 500, "tools/generated_tokens": 2517.314453125, "tools/num_python": 0.0, "tools/num_python_exec_error": 0.0, "tools/num_retrieval": 0.0, "tools/num_retriever_exec_error": 0.0, "tools/num_saving": 0.603515625, "tools/num_saving_exec_error": 0.0, "tools/num_tool_detect_error": 0.0 }, { "epoch": 0.35486160397445, "step": 500, "total_flos": 0.0, "train_loss": 0.04532187731936574, "train_runtime": 58434.4486, "train_samples_per_second": 4.381, "train_steps_per_second": 0.009 } ], "logging_steps": 1, "max_steps": 500, "num_input_tokens_seen": 1127757057, "num_train_epochs": 1, "save_steps": 100, "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": 8, "trial_name": null, "trial_params": null }