{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 7.905138339920948, "eval_steps": 500, "global_step": 2000, "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": 1.0, "completions/max_length": 114.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 49.45234375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.8, "completions/min_terminated_length": 0.0, "entropy": 1.0689914911985396, "epoch": 0.019762845849802372, "frac_reward_zero_std": 0.3, "grad_norm": 0.24201464653015137, "learning_rate": 8.88888888888889e-07, "loss": 0.115625, "num_tokens": 320739.0, "reward": -0.0315673828125, "reward/completion_length_max": 114.0, "reward/completion_length_mean": 49.45, "reward/completion_length_min": 27.8, "reward/completion_length_p90": 65.4, "reward/completion_length_p95": 86.0, "reward/duplicate_query_rate": 0.8984375, "reward/duplicate_ranked_plain_query_rate": 0.8984375, "reward/invalid_output_rate": 0.96015625, "reward/length_out_of_range_rate": 0.021875, "reward/low_length_query_rate": 0.021875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": -0.031640625, "reward/mean_colbert": 0.0386474609375, "reward/mean_mrr": 0.039453125, "reward/mean_ndcg_at_10": 0.0297119140625, "reward/mean_r_ndcg": 0.0248779296875, "reward/mean_r_ndcg_cutoff": 0.1984375, "reward/mean_recall": 0.00941162109375, "reward/mean_recall_at_10": 0.001953125, "reward/mean_recall_at_100": 0.00941162109375, "reward/mean_retrieval_combined_bounded": 0.01917724609375, "reward/mean_retrieval_combined_raw": 0.1572265625, "reward/output_tag_rate": 0.96015625, "reward/query_valid_rate": 0.03984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.03984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 2.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.021875, "reward/too_short_query_rate": 0.0, "reward_std": 0.48490562438964846, "rewards/retrieval_reward/mean": -0.031640625, "rewards/retrieval_reward/std": 0.48671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.842627763748169, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.7142605066299439, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 2.5895172119140626, "sampling/raw_vllm_sampling_logp_difference/mean": 0.07774874046444893, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 5, "step_time": 17.272163909766824 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 128.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 50.61484375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.9794655129313469, "epoch": 0.039525691699604744, "frac_reward_zero_std": 0.4, "grad_norm": 0.15731684863567352, "learning_rate": 2.0000000000000003e-06, "loss": 0.13828125, "num_tokens": 642806.0, "reward": -0.06875, "reward/completion_length_max": 128.0, "reward/completion_length_mean": 50.6, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 73.6, "reward/completion_length_p95": 77.6, "reward/duplicate_query_rate": 0.93515625, "reward/duplicate_ranked_plain_query_rate": 0.93515625, "reward/invalid_output_rate": 0.98828125, "reward/length_out_of_range_rate": 0.01328125, "reward/low_length_query_rate": 0.01328125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": -0.06875, "reward/mean_colbert": 0.01141357421875, "reward/mean_mrr": 0.01171875, "reward/mean_ndcg_at_10": 0.0104217529296875, "reward/mean_r_ndcg": 0.00950927734375, "reward/mean_r_ndcg_cutoff": 0.07578125, "reward/mean_recall": 0.00272979736328125, "reward/mean_recall_at_10": 0.0005626678466796875, "reward/mean_recall_at_100": 0.00272979736328125, "reward/mean_retrieval_combined_bounded": 0.005859375, "reward/mean_retrieval_combined_raw": 0.07900390625, "reward/output_tag_rate": 0.96015625, "reward/query_valid_rate": 0.01171875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.01171875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 7.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01328125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3937197089195251, "rewards/retrieval_reward/mean": -0.06875, "rewards/retrieval_reward/std": 0.39453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.894967031478882, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.7857137680053711, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 4.88275375366211, "sampling/raw_vllm_sampling_logp_difference/mean": 0.06589128598570823, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 10, "step_time": 16.811979734606574 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 128.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 48.2484375, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.8, "completions/min_terminated_length": 0.0, "entropy": 1.082428441941738, "epoch": 0.05928853754940711, "frac_reward_zero_std": 0.2, "grad_norm": 0.4188099503517151, "learning_rate": 3.1111111111111116e-06, "loss": 0.115625, "num_tokens": 961972.0, "reward": 0.03795166015625, "reward/completion_length_max": 128.0, "reward/completion_length_mean": 48.3, "reward/completion_length_min": 26.8, "reward/completion_length_p90": 60.8, "reward/completion_length_p95": 64.8, "reward/duplicate_query_rate": 0.9046875, "reward/duplicate_ranked_plain_query_rate": 0.9046875, "reward/invalid_output_rate": 0.94375, "reward/length_out_of_range_rate": 0.021875, "reward/low_length_query_rate": 0.021875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.0378662109375, "reward/mean_colbert": 0.0545166015625, "reward/mean_mrr": 0.05625, "reward/mean_ndcg_at_10": 0.0453369140625, "reward/mean_r_ndcg": 0.032763671875, "reward/mean_r_ndcg_cutoff": 0.384375, "reward/mean_recall": 0.0128387451171875, "reward/mean_recall_at_10": 0.00273284912109375, "reward/mean_recall_at_100": 0.0128387451171875, "reward/mean_retrieval_combined_bounded": 0.02701416015625, "reward/mean_retrieval_combined_raw": 0.18564453125, "reward/output_tag_rate": 0.9828125, "reward/query_valid_rate": 0.05625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.05625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 12.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.02109375, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.4200014531612396, "rewards/retrieval_reward/mean": 0.037890625, "rewards/retrieval_reward/std": 0.420703125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.804853057861328, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.721617865562439, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 3.3113525390625, "sampling/raw_vllm_sampling_logp_difference/mean": 0.07726039141416549, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 15, "step_time": 16.918399897939526 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 128.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 50.23359375, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.6, "completions/min_terminated_length": 0.0, "entropy": 0.9721478804945946, "epoch": 0.07905138339920949, "frac_reward_zero_std": 0.225, "grad_norm": 0.3254030644893646, "learning_rate": 4.222222222222223e-06, "loss": 0.140625, "num_tokens": 1283935.0, "reward": 0.001123046875, "reward/completion_length_max": 128.0, "reward/completion_length_mean": 50.2, "reward/completion_length_min": 26.6, "reward/completion_length_p90": 64.6, "reward/completion_length_p95": 91.6, "reward/duplicate_query_rate": 0.84140625, "reward/duplicate_ranked_plain_query_rate": 0.84140625, "reward/invalid_output_rate": 0.925, "reward/length_out_of_range_rate": 0.03828125, "reward/low_length_query_rate": 0.03828125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.00107421875, "reward/mean_colbert": 0.072607421875, "reward/mean_mrr": 0.075, "reward/mean_ndcg_at_10": 0.060986328125, "reward/mean_r_ndcg": 0.0467041015625, "reward/mean_r_ndcg_cutoff": 0.39140625, "reward/mean_recall": 0.0136474609375, "reward/mean_recall_at_10": 0.0034271240234375, "reward/mean_recall_at_100": 0.0136474609375, "reward/mean_retrieval_combined_bounded": 0.03583984375, "reward/mean_retrieval_combined_raw": 0.204296875, "reward/output_tag_rate": 0.9546875, "reward/query_valid_rate": 0.075, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.075, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 17.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0375, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.5855522572994232, "rewards/retrieval_reward/mean": 0.001123046875, "rewards/retrieval_reward/std": 0.5875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.9149232864379884, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.7506334781646729, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 3.956116485595703, "sampling/raw_vllm_sampling_logp_difference/mean": 0.06765746921300889, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 20, "step_time": 17.00318091964582 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 128.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 46.5875, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.6, "completions/min_terminated_length": 0.0, "entropy": 1.02460478246212, "epoch": 0.09881422924901186, "frac_reward_zero_std": 0.375, "grad_norm": 0.2704702615737915, "learning_rate": 5.333333333333334e-06, "loss": 0.076171875, "num_tokens": 1600815.0, "reward": 0.058551025390625, "reward/completion_length_max": 128.0, "reward/completion_length_mean": 46.55, "reward/completion_length_min": 25.6, "reward/completion_length_p90": 59.0, "reward/completion_length_p95": 64.0, "reward/duplicate_query_rate": 0.89921875, "reward/duplicate_ranked_plain_query_rate": 0.89921875, "reward/invalid_output_rate": 0.93515625, "reward/length_out_of_range_rate": 0.0234375, "reward/low_length_query_rate": 0.0234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.05859375, "reward/mean_colbert": 0.062713623046875, "reward/mean_mrr": 0.06357421875, "reward/mean_ndcg_at_10": 0.057049560546875, "reward/mean_r_ndcg": 0.047784423828125, "reward/mean_r_ndcg_cutoff": 0.38046875, "reward/mean_recall": 0.0112762451171875, "reward/mean_recall_at_10": 0.0030231475830078125, "reward/mean_recall_at_100": 0.0112762451171875, "reward/mean_retrieval_combined_bounded": 0.030419921875, "reward/mean_retrieval_combined_raw": 0.18828125, "reward/output_tag_rate": 0.9875, "reward/query_valid_rate": 0.06484375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.06484375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 22.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.02109375, "reward/too_short_query_rate": 0.00234375, "reward_std": 0.42350157499313357, "rewards/retrieval_reward/mean": 0.05849609375, "rewards/retrieval_reward/std": 0.423828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.860168218612671, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.7259661912918091, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 2.6333358764648436, "sampling/raw_vllm_sampling_logp_difference/mean": 0.07660264670848846, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 25, "step_time": 16.805487859807908 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.9984375, "completions/max_length": 128.0, "completions/max_terminated_length": 17.8, "completions/mean_length": 46.94375, "completions/mean_terminated_length": 17.8, "completions/min_length": 27.0, "completions/min_terminated_length": 17.8, "entropy": 1.000285193324089, "epoch": 0.11857707509881422, "frac_reward_zero_std": 0.2, "grad_norm": 0.4384768307209015, "learning_rate": 6.444444444444445e-06, "loss": 0.11953125, "num_tokens": 1918375.0, "reward": 0.035986328125, "reward/completion_length_max": 128.0, "reward/completion_length_mean": 47.0, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 59.8, "reward/completion_length_p95": 66.0, "reward/duplicate_query_rate": 0.875, "reward/duplicate_ranked_plain_query_rate": 0.875, "reward/invalid_output_rate": 0.93046875, "reward/length_out_of_range_rate": 0.03046875, "reward/low_length_query_rate": 0.03046875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.03594970703125, "reward/mean_colbert": 0.067626953125, "reward/mean_mrr": 0.064453125, "reward/mean_ndcg_at_10": 0.051611328125, "reward/mean_r_ndcg": 0.0412109375, "reward/mean_r_ndcg_cutoff": 0.34453125, "reward/mean_recall": 0.0164306640625, "reward/mean_recall_at_10": 0.002886962890625, "reward/mean_recall_at_100": 0.0164306640625, "reward/mean_retrieval_combined_bounded": 0.03173828125, "reward/mean_retrieval_combined_raw": 0.2166015625, "reward/output_tag_rate": 0.975, "reward/query_valid_rate": 0.06953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.06953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 27.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.03046875, "reward/too_short_query_rate": 0.0, "reward_std": 0.49257102608680725, "rewards/retrieval_reward/mean": 0.035931396484375, "rewards/retrieval_reward/std": 0.493359375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.9375321865081787, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.7365406513214111, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 5.537294387817383, "sampling/raw_vllm_sampling_logp_difference/mean": 0.08235383927822112, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 30, "step_time": 16.851179276558106 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 118.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 45.965625, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.6, "completions/min_terminated_length": 0.0, "entropy": 0.9591664433479309, "epoch": 0.1383399209486166, "frac_reward_zero_std": 0.275, "grad_norm": 0.27477797865867615, "learning_rate": 7.555555555555556e-06, "loss": 0.093359375, "num_tokens": 2235067.0, "reward": 0.05771484375, "reward/completion_length_max": 118.8, "reward/completion_length_mean": 46.0, "reward/completion_length_min": 26.6, "reward/completion_length_p90": 59.8, "reward/completion_length_p95": 65.0, "reward/duplicate_query_rate": 0.8984375, "reward/duplicate_ranked_plain_query_rate": 0.8984375, "reward/invalid_output_rate": 0.934375, "reward/length_out_of_range_rate": 0.02109375, "reward/low_length_query_rate": 0.02109375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.057666015625, "reward/mean_colbert": 0.063671875, "reward/mean_mrr": 0.063818359375, "reward/mean_ndcg_at_10": 0.0507080078125, "reward/mean_r_ndcg": 0.03992919921875, "reward/mean_r_ndcg_cutoff": 0.37890625, "reward/mean_recall": 0.0136962890625, "reward/mean_recall_at_10": 0.00298309326171875, "reward/mean_recall_at_100": 0.0136962890625, "reward/mean_retrieval_combined_bounded": 0.0322998046875, "reward/mean_retrieval_combined_raw": 0.242578125, "reward/output_tag_rate": 0.98515625, "reward/query_valid_rate": 0.065625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.065625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 32.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.02109375, "reward/too_short_query_rate": 0.0, "reward_std": 0.4418700456619263, "rewards/retrieval_reward/mean": 0.05771484375, "rewards/retrieval_reward/std": 0.442578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.9157299518585207, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.729906690120697, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 2.362660789489746, "sampling/raw_vllm_sampling_logp_difference/mean": 0.07508442252874374, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 35, "step_time": 16.431749085756017 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 104.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 39.71484375, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.6, "completions/min_terminated_length": 0.0, "entropy": 0.942545510828495, "epoch": 0.15810276679841898, "frac_reward_zero_std": 0.15, "grad_norm": 0.5535351634025574, "learning_rate": 8.666666666666668e-06, "loss": 0.055078125, "num_tokens": 2543758.0, "reward": 0.260791015625, "reward/completion_length_max": 104.6, "reward/completion_length_mean": 39.7, "reward/completion_length_min": 25.6, "reward/completion_length_p90": 49.2, "reward/completion_length_p95": 53.4, "reward/duplicate_query_rate": 0.765625, "reward/duplicate_ranked_plain_query_rate": 0.765625, "reward/invalid_output_rate": 0.8015625, "reward/length_out_of_range_rate": 0.02890625, "reward/low_length_query_rate": 0.02890625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.2609375, "reward/mean_colbert": 0.19296875, "reward/mean_mrr": 0.198046875, "reward/mean_ndcg_at_10": 0.16357421875, "reward/mean_r_ndcg": 0.1294921875, "reward/mean_r_ndcg_cutoff": 1.47890625, "reward/mean_recall": 0.046630859375, "reward/mean_recall_at_10": 0.009747314453125, "reward/mean_recall_at_100": 0.046630859375, "reward/mean_retrieval_combined_bounded": 0.09072265625, "reward/mean_retrieval_combined_raw": 0.403125, "reward/output_tag_rate": 0.99296875, "reward/query_valid_rate": 0.1984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.1984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 37.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0265625, "reward/too_short_query_rate": 0.00234375, "reward_std": 0.6178877115249634, "rewards/retrieval_reward/mean": 0.2611328125, "rewards/retrieval_reward/std": 0.61875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8712701320648195, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.7800331234931945, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 2.5099894523620607, "sampling/raw_vllm_sampling_logp_difference/mean": 0.07183510959148406, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 40, "step_time": 15.920656395808328 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 116.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 38.9515625, "completions/mean_terminated_length": 0.0, "completions/min_length": 23.2, "completions/min_terminated_length": 0.0, "entropy": 0.8701661348342895, "epoch": 0.17786561264822134, "frac_reward_zero_std": 0.1, "grad_norm": 0.39455440640449524, "learning_rate": 9.777777777777779e-06, "loss": 0.06640625, "num_tokens": 2851472.0, "reward": 0.23641357421875, "reward/completion_length_max": 116.0, "reward/completion_length_mean": 38.95, "reward/completion_length_min": 23.2, "reward/completion_length_p90": 49.0, "reward/completion_length_p95": 52.8, "reward/duplicate_query_rate": 0.7484375, "reward/duplicate_ranked_plain_query_rate": 0.7484375, "reward/invalid_output_rate": 0.809375, "reward/length_out_of_range_rate": 0.0546875, "reward/low_length_query_rate": 0.0546875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.2361328125, "reward/mean_colbert": 0.1857421875, "reward/mean_mrr": 0.190625, "reward/mean_ndcg_at_10": 0.1458984375, "reward/mean_r_ndcg": 0.1064453125, "reward/mean_r_ndcg_cutoff": 1.17734375, "reward/mean_recall": 0.047412109375, "reward/mean_recall_at_10": 0.00916748046875, "reward/mean_recall_at_100": 0.047412109375, "reward/mean_retrieval_combined_bounded": 0.085546875, "reward/mean_retrieval_combined_raw": 0.4109375, "reward/output_tag_rate": 0.99375, "reward/query_valid_rate": 0.190625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.190625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 42.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.04296875, "reward/too_short_query_rate": 0.01171875, "reward_std": 0.6137177467346191, "rewards/retrieval_reward/mean": 0.236328125, "rewards/retrieval_reward/std": 0.61484375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8813183307647705, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.7700004935264587, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 2.211732339859009, "sampling/raw_vllm_sampling_logp_difference/mean": 0.07570562660694122, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 45, "step_time": 16.415317437390332 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 100.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.12890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 22.4, "completions/min_terminated_length": 0.0, "entropy": 0.7714573800563812, "epoch": 0.1976284584980237, "frac_reward_zero_std": 0.125, "grad_norm": 0.5634564757347107, "learning_rate": 1.088888888888889e-05, "loss": 0.07890625, "num_tokens": 3155125.0, "reward": 0.3758056640625, "reward/completion_length_max": 100.6, "reward/completion_length_mean": 36.1, "reward/completion_length_min": 22.4, "reward/completion_length_p90": 45.8, "reward/completion_length_p95": 49.4, "reward/duplicate_query_rate": 0.625, "reward/duplicate_ranked_plain_query_rate": 0.625, "reward/invalid_output_rate": 0.69765625, "reward/length_out_of_range_rate": 0.065625, "reward/low_length_query_rate": 0.065625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.3759765625, "reward/mean_colbert": 0.2947265625, "reward/mean_mrr": 0.3015625, "reward/mean_ndcg_at_10": 0.268359375, "reward/mean_r_ndcg": 0.22109375, "reward/mean_r_ndcg_cutoff": 2.265625, "reward/mean_recall": 0.063623046875, "reward/mean_recall_at_10": 0.01455078125, "reward/mean_recall_at_100": 0.063623046875, "reward/mean_retrieval_combined_bounded": 0.12041015625, "reward/mean_retrieval_combined_raw": 0.479296875, "reward/output_tag_rate": 0.99296875, "reward/query_valid_rate": 0.30234375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.30234375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 47.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0421875, "reward/too_short_query_rate": 0.0234375, "reward_std": 0.6885141730308533, "rewards/retrieval_reward/mean": 0.3759765625, "rewards/retrieval_reward/std": 0.68984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7564703941345217, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.4978547692298889, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 3.3505670547485353, "sampling/raw_vllm_sampling_logp_difference/mean": 0.1130765751004219, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 50, "step_time": 15.835850008798298 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 46.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.3046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 20.0, "completions/min_terminated_length": 0.0, "entropy": 0.6783149242401123, "epoch": 0.2015810276679842, "frac_reward_zero_std": 0.0, "num_tokens": 3215043.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 3.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1135.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.5940604881750182, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.6507943444369686, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.852863436123348, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9938325991189427, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6097818356885268, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6640569018171076, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8533333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 2.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 413.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3233269449747422, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.36332143165072095, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5351089588377724, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8619854721549637, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 4.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 280.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.4595827694421934, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.4750864577856444, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.6143529684601112, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.8942474489795919, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 10.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 313.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5818003216749389, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.365889615027559, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.18023520978590576, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3468554409088048, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 32.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3420.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.3373876271603708, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.3702210536340771, "retrieval_eval/nq/single_turn/recall_at_10": 0.5571881091617934, "retrieval_eval/nq/single_turn/recall_at_100": 0.8498294346978557, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 1.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 299.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6603984559485794, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6973493096979007, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8525641025641025, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9449832775919733, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 0.545623779296875, "reward/completion_length_max": 46.0, "reward/completion_length_mean": 32.25, "reward/completion_length_min": 20.0, "reward/completion_length_p90": 39.0, "reward/completion_length_p95": 41.0, "reward/duplicate_query_rate": 0.51171875, "reward/duplicate_ranked_plain_query_rate": 0.51171875, "reward/invalid_output_rate": 0.5859375, "reward/length_out_of_range_rate": 0.07421875, "reward/low_length_query_rate": 0.07421875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.546875, "reward/mean_colbert": 0.400390625, "reward/mean_mrr": 0.4140625, "reward/mean_ndcg_at_10": 0.328125, "reward/mean_r_ndcg": 0.25390625, "reward/mean_r_ndcg_cutoff": 2.296875, "reward/mean_recall": 0.1328125, "reward/mean_recall_at_10": 0.020263671875, "reward/mean_recall_at_100": 0.1328125, "reward/mean_retrieval_combined_bounded": 0.1689453125, "reward/mean_retrieval_combined_raw": 0.609375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.4140625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.4140625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 50.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.02734375, "reward/too_short_query_rate": 0.046875, "reward_std": 0.7390519380569458, "rewards/retrieval_reward/mean": 0.546875, "rewards/retrieval_reward/std": 0.73828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4581339359283447, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21889740228652954, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 3.872495651245117, "sampling/raw_vllm_sampling_logp_difference/mean": 0.16708062589168549, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 51, "step_time": 13.281792772992048 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 50.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.4013671875, "completions/mean_terminated_length": 0.0, "completions/min_length": 20.25, "completions/min_terminated_length": 0.0, "entropy": 0.6522187441587448, "epoch": 0.21739130434782608, "frac_reward_zero_std": 0.0625, "grad_norm": 0.7386945486068726, "learning_rate": 1.2e-05, "loss": 0.0111328125, "num_tokens": 3452222.0, "reward": 0.4326744079589844, "reward/completion_length_max": 50.0, "reward/completion_length_mean": 30.375, "reward/completion_length_min": 20.25, "reward/completion_length_p90": 39.0, "reward/completion_length_p95": 41.5, "reward/duplicate_query_rate": 0.4599609375, "reward/duplicate_ranked_plain_query_rate": 0.4599609375, "reward/invalid_output_rate": 0.623046875, "reward/length_out_of_range_rate": 0.1630859375, "reward/low_length_query_rate": 0.1630859375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.43212890625, "reward/mean_colbert": 0.3662109375, "reward/mean_mrr": 0.3642578125, "reward/mean_ndcg_at_10": 0.304443359375, "reward/mean_r_ndcg": 0.222412109375, "reward/mean_r_ndcg_cutoff": 2.861328125, "reward/mean_recall": 0.0819091796875, "reward/mean_recall_at_10": 0.017547607421875, "reward/mean_recall_at_100": 0.0819091796875, "reward/mean_retrieval_combined_bounded": 0.1370849609375, "reward/mean_retrieval_combined_raw": 0.5224609375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.376953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.376953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 52.5, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0205078125, "reward/too_short_query_rate": 0.142578125, "reward_std": 0.7452941834926605, "rewards/retrieval_reward/mean": 0.43212890625, "rewards/retrieval_reward/std": 0.748046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6444643139839172, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2182563915848732, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 4.93527364730835, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20653125271201134, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 55, "step_time": 13.381144469982246 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 79.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.74921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 19.4, "completions/min_terminated_length": 0.0, "entropy": 0.6033129513263702, "epoch": 0.23715415019762845, "frac_reward_zero_std": 0.0, "grad_norm": 0.5191141963005066, "learning_rate": 1.3111111111111113e-05, "loss": 0.0421875, "num_tokens": 3746365.0, "reward": 0.4834716796875, "reward/completion_length_max": 79.2, "reward/completion_length_mean": 28.775, "reward/completion_length_min": 19.4, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 37.4, "reward/duplicate_query_rate": 0.290625, "reward/duplicate_ranked_plain_query_rate": 0.290625, "reward/invalid_output_rate": 0.54375, "reward/length_out_of_range_rate": 0.246875, "reward/low_length_query_rate": 0.246875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.48359375, "reward/mean_colbert": 0.44375, "reward/mean_mrr": 0.4501953125, "reward/mean_ndcg_at_10": 0.36953125, "reward/mean_r_ndcg": 0.3056640625, "reward/mean_r_ndcg_cutoff": 2.984375, "reward/mean_recall": 0.09599609375, "reward/mean_recall_at_10": 0.0218017578125, "reward/mean_recall_at_100": 0.09599609375, "reward/mean_retrieval_combined_bounded": 0.16318359375, "reward/mean_retrieval_combined_raw": 0.555859375, "reward/output_tag_rate": 0.99375, "reward/query_valid_rate": 0.45625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.45625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 57.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01171875, "reward/too_short_query_rate": 0.23515625, "reward_std": 0.8085848331451416, "rewards/retrieval_reward/mean": 0.48359375, "rewards/retrieval_reward/std": 0.8078125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7086087226867677, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18895550668239594, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 5.632321166992187, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2388474762439728, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 60, "step_time": 14.910125167027582 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 44.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.88515625, "completions/mean_terminated_length": 0.0, "completions/min_length": 21.0, "completions/min_terminated_length": 0.0, "entropy": 0.6603240117430687, "epoch": 0.25691699604743085, "frac_reward_zero_std": 0.05, "grad_norm": 0.7652059197425842, "learning_rate": 1.4222222222222224e-05, "loss": -0.0001007080078125, "num_tokens": 4040426.0, "reward": 0.56324462890625, "reward/completion_length_max": 44.8, "reward/completion_length_mean": 28.9, "reward/completion_length_min": 21.0, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 38.0, "reward/duplicate_query_rate": 0.3046875, "reward/duplicate_ranked_plain_query_rate": 0.3046875, "reward/invalid_output_rate": 0.49921875, "reward/length_out_of_range_rate": 0.19453125, "reward/low_length_query_rate": 0.19453125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.563671875, "reward/mean_colbert": 0.48828125, "reward/mean_mrr": 0.498828125, "reward/mean_ndcg_at_10": 0.4046875, "reward/mean_r_ndcg": 0.338671875, "reward/mean_r_ndcg_cutoff": 3.190625, "reward/mean_recall": 0.114453125, "reward/mean_recall_at_10": 0.0241455078125, "reward/mean_recall_at_100": 0.114453125, "reward/mean_retrieval_combined_bounded": 0.159765625, "reward/mean_retrieval_combined_raw": 0.52578125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.50078125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.50078125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 62.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.015625, "reward/too_short_query_rate": 0.17890625, "reward_std": 0.7646669149398804, "rewards/retrieval_reward/mean": 0.563671875, "rewards/retrieval_reward/std": 0.76640625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7909157276153564, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.22618365585803984, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 4.80064640045166, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23094754815101623, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 65, "step_time": 13.30590343303047 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 60.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.25625, "completions/mean_terminated_length": 0.0, "completions/min_length": 21.0, "completions/min_terminated_length": 0.0, "entropy": 0.7701092317700386, "epoch": 0.2766798418972332, "frac_reward_zero_std": 0.025, "grad_norm": 0.6565812230110168, "learning_rate": 1.5333333333333334e-05, "loss": 0.01806640625, "num_tokens": 4337906.0, "reward": 0.8337371826171875, "reward/completion_length_max": 60.6, "reward/completion_length_mean": 31.25, "reward/completion_length_min": 21.0, "reward/completion_length_p90": 37.4, "reward/completion_length_p95": 38.8, "reward/duplicate_query_rate": 0.22734375, "reward/duplicate_ranked_plain_query_rate": 0.22734375, "reward/invalid_output_rate": 0.29453125, "reward/length_out_of_range_rate": 0.06640625, "reward/low_length_query_rate": 0.06640625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.834375, "reward/mean_colbert": 0.6859375, "reward/mean_mrr": 0.7046875, "reward/mean_ndcg_at_10": 0.609375, "reward/mean_r_ndcg": 0.468359375, "reward/mean_r_ndcg_cutoff": 4.275, "reward/mean_recall": 0.17392578125, "reward/mean_recall_at_10": 0.0341064453125, "reward/mean_recall_at_100": 0.17392578125, "reward/mean_retrieval_combined_bounded": 0.162890625, "reward/mean_retrieval_combined_raw": 0.444140625, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.70546875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.70546875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 67.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01796875, "reward/too_short_query_rate": 0.0484375, "reward_std": 0.61112020611763, "rewards/retrieval_reward/mean": 0.834375, "rewards/retrieval_reward/std": 0.61171875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.926502990722656, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.35694006979465487, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 3.6481143951416017, "sampling/raw_vllm_sampling_logp_difference/mean": 0.1828453779220581, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 70, "step_time": 14.395688624202739 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 48.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.65234375, "completions/mean_terminated_length": 0.0, "completions/min_length": 22.6, "completions/min_terminated_length": 0.0, "entropy": 0.7925162807106971, "epoch": 0.2964426877470356, "frac_reward_zero_std": 0.025, "grad_norm": 0.7572252750396729, "learning_rate": 1.6444444444444444e-05, "loss": 0.0162109375, "num_tokens": 4635605.0, "reward": 0.853204345703125, "reward/completion_length_max": 48.0, "reward/completion_length_mean": 31.6, "reward/completion_length_min": 22.6, "reward/completion_length_p90": 37.0, "reward/completion_length_p95": 38.6, "reward/duplicate_query_rate": 0.23828125, "reward/duplicate_ranked_plain_query_rate": 0.23828125, "reward/invalid_output_rate": 0.2875, "reward/length_out_of_range_rate": 0.0484375, "reward/low_length_query_rate": 0.0484375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.85390625, "reward/mean_colbert": 0.69140625, "reward/mean_mrr": 0.69765625, "reward/mean_ndcg_at_10": 0.549609375, "reward/mean_r_ndcg": 0.42109375, "reward/mean_r_ndcg_cutoff": 3.815625, "reward/mean_recall": 0.174609375, "reward/mean_recall_at_10": 0.0329345703125, "reward/mean_recall_at_100": 0.174609375, "reward/mean_retrieval_combined_bounded": 0.16640625, "reward/mean_retrieval_combined_raw": 0.470703125, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.7125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.7125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 72.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0203125, "reward/too_short_query_rate": 0.028125, "reward_std": 0.6168259620666504, "rewards/retrieval_reward/mean": 0.85390625, "rewards/retrieval_reward/std": 0.61875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7351630687713624, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.31711265444755554, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 5.460562515258789, "sampling/raw_vllm_sampling_logp_difference/mean": 0.18884131610393523, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 75, "step_time": 13.831003672559746 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 44.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.49921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 23.4, "completions/min_terminated_length": 0.0, "entropy": 0.8097933113574982, "epoch": 0.31620553359683795, "frac_reward_zero_std": 0.025, "grad_norm": 0.6834232211112976, "learning_rate": 1.7555555555555556e-05, "loss": 0.00791015625, "num_tokens": 4933140.0, "reward": 0.96724853515625, "reward/completion_length_max": 44.6, "reward/completion_length_mean": 31.525, "reward/completion_length_min": 23.4, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 38.2, "reward/duplicate_query_rate": 0.1296875, "reward/duplicate_ranked_plain_query_rate": 0.1296875, "reward/invalid_output_rate": 0.159375, "reward/length_out_of_range_rate": 0.02890625, "reward/low_length_query_rate": 0.02890625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.96796875, "reward/mean_colbert": 0.81796875, "reward/mean_mrr": 0.815625, "reward/mean_ndcg_at_10": 0.6703125, "reward/mean_r_ndcg": 0.5546875, "reward/mean_r_ndcg_cutoff": 6.4625, "reward/mean_recall": 0.190625, "reward/mean_recall_at_10": 0.03984375, "reward/mean_recall_at_100": 0.190625, "reward/mean_retrieval_combined_bounded": 0.14287109375, "reward/mean_retrieval_combined_raw": 0.345703125, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.840625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.840625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 77.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01171875, "reward/too_short_query_rate": 0.0171875, "reward_std": 0.5096926033496857, "rewards/retrieval_reward/mean": 0.96796875, "rewards/retrieval_reward/std": 0.509765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.9100707530975343, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2940949261188507, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.490417098999023, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20723369717597961, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 80, "step_time": 13.81961263458943 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.89296875, "completions/mean_terminated_length": 0.0, "completions/min_length": 23.2, "completions/min_terminated_length": 0.0, "entropy": 0.7922001481056213, "epoch": 0.3359683794466403, "frac_reward_zero_std": 0.0, "grad_norm": 0.8614085912704468, "learning_rate": 1.866666666666667e-05, "loss": 0.00439453125, "num_tokens": 5228939.0, "reward": 0.9767242431640625, "reward/completion_length_max": 41.6, "reward/completion_length_mean": 29.925, "reward/completion_length_min": 23.2, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.0390625, "reward/duplicate_ranked_plain_query_rate": 0.0390625, "reward/invalid_output_rate": 0.10078125, "reward/length_out_of_range_rate": 0.06171875, "reward/low_length_query_rate": 0.06171875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.97578125, "reward/mean_colbert": 0.87578125, "reward/mean_mrr": 0.8984375, "reward/mean_ndcg_at_10": 0.7421875, "reward/mean_r_ndcg": 0.5765625, "reward/mean_r_ndcg_cutoff": 5.08125, "reward/mean_recall": 0.1974609375, "reward/mean_recall_at_10": 0.04365234375, "reward/mean_recall_at_100": 0.1974609375, "reward/mean_retrieval_combined_bounded": 0.10810546875, "reward/mean_retrieval_combined_raw": 0.26181640625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.89921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.89921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 82.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01875, "reward/too_short_query_rate": 0.04296875, "reward_std": 0.4613050639629364, "rewards/retrieval_reward/mean": 0.97578125, "rewards/retrieval_reward/std": 0.4625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8694451808929444, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.31136937737464904, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 5.842010879516602, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21768843233585358, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 85, "step_time": 13.907765987014864 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.98046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 22.6, "completions/min_terminated_length": 0.0, "entropy": 0.7338098272681236, "epoch": 0.3557312252964427, "frac_reward_zero_std": 0.0, "grad_norm": 1.2660411596298218, "learning_rate": 1.977777777777778e-05, "loss": -0.010986328125, "num_tokens": 5523570.0, "reward": 0.9992584228515625, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 29.0, "reward/completion_length_min": 22.6, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 34.2, "reward/duplicate_query_rate": 0.0265625, "reward/duplicate_ranked_plain_query_rate": 0.0265625, "reward/invalid_output_rate": 0.06484375, "reward/length_out_of_range_rate": 0.03828125, "reward/low_length_query_rate": 0.03828125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99921875, "reward/mean_colbert": 0.91171875, "reward/mean_mrr": 0.93203125, "reward/mean_ndcg_at_10": 0.75703125, "reward/mean_r_ndcg": 0.571875, "reward/mean_r_ndcg_cutoff": 6.01875, "reward/mean_recall": 0.2439453125, "reward/mean_recall_at_10": 0.04482421875, "reward/mean_recall_at_100": 0.2439453125, "reward/mean_retrieval_combined_bounded": 0.083203125, "reward/mean_retrieval_combined_raw": 0.19296875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.93515625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.93515625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 87.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0375, "reward_std": 0.42105095982551577, "rewards/retrieval_reward/mean": 0.99921875, "rewards/retrieval_reward/std": 0.421875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8669458866119384, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2414372444152832, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.509513092041016, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2560146689414978, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 90, "step_time": 13.75365567955887 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.778125, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.2, "completions/min_terminated_length": 0.0, "entropy": 0.7689405232667923, "epoch": 0.37549407114624506, "frac_reward_zero_std": 0.0, "grad_norm": 1.1739084720611572, "learning_rate": 2e-05, "loss": 0.005224609375, "num_tokens": 5820630.0, "reward": 1.017724609375, "reward/completion_length_max": 40.8, "reward/completion_length_mean": 30.8, "reward/completion_length_min": 25.2, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.01640625, "reward/duplicate_ranked_plain_query_rate": 0.01640625, "reward/invalid_output_rate": 0.0203125, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.95234375, "reward/mean_mrr": 0.978125, "reward/mean_ndcg_at_10": 0.7984375, "reward/mean_r_ndcg": 0.565625, "reward/mean_r_ndcg_cutoff": 5.75625, "reward/mean_recall": 0.2177734375, "reward/mean_recall_at_10": 0.04716796875, "reward/mean_recall_at_100": 0.2177734375, "reward/mean_retrieval_combined_bounded": 0.04013671875, "reward/mean_retrieval_combined_raw": 0.08984375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9796875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9796875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 92.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.003125, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.3185440480709076, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.319140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.541017484664917, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20860912203788756, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 4.899068450927734, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2568970173597336, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 95, "step_time": 13.924750573188067 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 42.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.75390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.2, "completions/min_terminated_length": 0.0, "entropy": 0.7354949817061425, "epoch": 0.3952569169960474, "frac_reward_zero_std": 0.0, "grad_norm": 1.0500935316085815, "learning_rate": 2e-05, "loss": 0.0041748046875, "num_tokens": 6117595.0, "reward": 1.0330230712890625, "reward/completion_length_max": 42.0, "reward/completion_length_mean": 30.775, "reward/completion_length_min": 25.2, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 35.4, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.0140625, "reward/length_out_of_range_rate": 0.00859375, "reward/low_length_query_rate": 0.00859375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.959375, "reward/mean_mrr": 0.9859375, "reward/mean_ndcg_at_10": 0.88359375, "reward/mean_r_ndcg": 0.696875, "reward/mean_r_ndcg_cutoff": 5.95, "reward/mean_recall": 0.2201171875, "reward/mean_recall_at_10": 0.048486328125, "reward/mean_recall_at_100": 0.2201171875, "reward/mean_retrieval_combined_bounded": 0.05126953125, "reward/mean_retrieval_combined_raw": 0.1412109375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9859375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9859375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 97.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00703125, "reward/too_short_query_rate": 0.0015625, "reward_std": 0.3148643672466278, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.315625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8388492584228517, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18410162925720214, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.330235481262207, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2935036808252335, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 100, "step_time": 13.995502258802299 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.47265625, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.0, "completions/min_terminated_length": 0.0, "entropy": 0.6828476786613464, "epoch": 0.39920948616600793, "frac_reward_zero_std": 0.0, "num_tokens": 6176564.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.5956358920887923, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.6511986197213948, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8506151142355008, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9964850615114236, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5563651656004013, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6121497448993545, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8066666666666666, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.98, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.31165287484468285, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.35273154042506666, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5277108433734939, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8578313253012049, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.5637722964711364, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.5818582310774283, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7250567800743858, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9249958081824279, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5602414280532507, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.35338559262183933, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.1819996858381748, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3301435433839657, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 0.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3452.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.32376401996878473, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.3561803864760306, "retrieval_eval/nq/single_turn/recall_at_10": 0.5420529161838548, "retrieval_eval/nq/single_turn/recall_at_100": 0.8227356122054847, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6559106931678991, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7016665377743952, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8684444444444446, "retrieval_eval/scifact/single_turn/recall_at_100": 0.95, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0082550048828125, "reward/completion_length_max": 37.0, "reward/completion_length_mean": 29.5, "reward/completion_length_min": 24.0, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0078125, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.78125, "reward/mean_r_ndcg": 0.63671875, "reward/mean_r_ndcg_cutoff": 8.0625, "reward/mean_recall": 0.24609375, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.24609375, "reward/mean_retrieval_combined_bounded": 0.01422119140625, "reward/mean_retrieval_combined_raw": 0.0262451171875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 100.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00390625, "reward_std": 0.3227255940437317, "rewards/retrieval_reward/mean": 1.0078125, "rewards/retrieval_reward/std": 0.32421875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.986923933029175, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.25042882561683655, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.906366348266602, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2543919086456299, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 101, "step_time": 13.666364710021298 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.7783203125, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.75, "completions/min_terminated_length": 0.0, "entropy": 0.6534224972128868, "epoch": 0.4150197628458498, "frac_reward_zero_std": 0.0, "grad_norm": 1.2430715560913086, "learning_rate": 2e-05, "loss": -0.00517578125, "num_tokens": 6412849.0, "reward": 1.0233039855957031, "reward/completion_length_max": 37.25, "reward/completion_length_mean": 29.8125, "reward/completion_length_min": 24.75, "reward/completion_length_p90": 32.75, "reward/completion_length_p95": 33.75, "reward/duplicate_query_rate": 0.005859375, "reward/duplicate_ranked_plain_query_rate": 0.005859375, "reward/invalid_output_rate": 0.009765625, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021484375, "reward/mean_colbert": 0.9658203125, "reward/mean_mrr": 0.990234375, "reward/mean_ndcg_at_10": 0.8466796875, "reward/mean_r_ndcg": 0.6640625, "reward/mean_r_ndcg_cutoff": 6.0078125, "reward/mean_recall": 0.2265625, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.2265625, "reward/mean_retrieval_combined_bounded": 0.0350341796875, "reward/mean_retrieval_combined_raw": 0.063385009765625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990234375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990234375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 102.5, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00390625, "reward_std": 0.29485099017620087, "rewards/retrieval_reward/mean": 1.021484375, "rewards/retrieval_reward/std": 0.2958984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.577809512615204, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12123176828026772, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.966328144073486, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3486843854188919, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 105, "step_time": 13.773067360772984 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.82890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.8, "completions/min_terminated_length": 0.0, "entropy": 0.7359675601124763, "epoch": 0.43478260869565216, "frac_reward_zero_std": 0.0, "grad_norm": 1.4177250862121582, "learning_rate": 2e-05, "loss": 0.003173828125, "num_tokens": 6709782.0, "reward": 1.005792236328125, "reward/completion_length_max": 39.4, "reward/completion_length_mean": 30.825, "reward/completion_length_min": 25.8, "reward/completion_length_p90": 33.8, "reward/completion_length_p95": 35.4, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.01640625, "reward/length_out_of_range_rate": 0.0140625, "reward/low_length_query_rate": 0.0140625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00625, "reward/mean_colbert": 0.9578125, "reward/mean_mrr": 0.9828125, "reward/mean_ndcg_at_10": 0.840625, "reward/mean_r_ndcg": 0.6734375, "reward/mean_r_ndcg_cutoff": 5.8625, "reward/mean_recall": 0.234765625, "reward/mean_recall_at_10": 0.047509765625, "reward/mean_recall_at_100": 0.234765625, "reward/mean_retrieval_combined_bounded": 0.02930908203125, "reward/mean_retrieval_combined_raw": 0.058251829072833064, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98359375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98359375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 107.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01328125, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.320960259437561, "rewards/retrieval_reward/mean": 1.00625, "rewards/retrieval_reward/std": 0.321875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.521856451034546, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.10245464891195297, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.161977386474609, "sampling/raw_vllm_sampling_logp_difference/mean": 0.35497636795043946, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 110, "step_time": 13.82557757444447 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.8859375, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.8, "completions/min_terminated_length": 0.0, "entropy": 0.7074582532048226, "epoch": 0.45454545454545453, "frac_reward_zero_std": 0.0, "grad_norm": 1.399561882019043, "learning_rate": 2e-05, "loss": -0.00022125244140625, "num_tokens": 7006820.0, "reward": 1.0124908447265626, "reward/completion_length_max": 38.8, "reward/completion_length_mean": 30.875, "reward/completion_length_min": 25.8, "reward/completion_length_p90": 33.8, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00546875, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0125, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.82734375, "reward/mean_r_ndcg": 0.68515625, "reward/mean_r_ndcg_cutoff": 6.41875, "reward/mean_recall": 0.223828125, "reward/mean_recall_at_10": 0.047802734375, "reward/mean_recall_at_100": 0.223828125, "reward/mean_retrieval_combined_bounded": 0.018670654296875, "reward/mean_retrieval_combined_raw": 0.02761194035410881, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99453125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99453125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 112.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.2920322120189667, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.292578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.162398862838745, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.06291390657424926, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.841758728027344, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4077290415763855, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 115, "step_time": 13.821823750960174 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.40546875, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.8, "completions/min_terminated_length": 0.0, "entropy": 0.6483606472611427, "epoch": 0.4743083003952569, "frac_reward_zero_std": 0.0, "grad_norm": 1.2174831628799438, "learning_rate": 2e-05, "loss": -0.0010986328125, "num_tokens": 7303083.0, "reward": 1.0282012939453125, "reward/completion_length_max": 37.2, "reward/completion_length_mean": 30.4, "reward/completion_length_min": 24.8, "reward/completion_length_p90": 33.4, "reward/completion_length_p95": 34.4, "reward/duplicate_query_rate": 0.009375, "reward/duplicate_ranked_plain_query_rate": 0.009375, "reward/invalid_output_rate": 0.01015625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.02890625, "reward/mean_colbert": 0.96640625, "reward/mean_mrr": 0.9890625, "reward/mean_ndcg_at_10": 0.84296875, "reward/mean_r_ndcg": 0.70859375, "reward/mean_r_ndcg_cutoff": 6.1125, "reward/mean_recall": 0.208984375, "reward/mean_recall_at_10": 0.047705078125, "reward/mean_recall_at_100": 0.208984375, "reward/mean_retrieval_combined_bounded": 0.0387481689453125, "reward/mean_retrieval_combined_raw": 0.07031329590827226, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 117.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.3020919859409332, "rewards/retrieval_reward/mean": 1.02890625, "rewards/retrieval_reward/std": 0.302734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.321732687950134, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.04760202094912529, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.345492172241212, "sampling/raw_vllm_sampling_logp_difference/mean": 0.48681633472442626, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 120, "step_time": 13.71840160281863 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.27421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.0, "completions/min_terminated_length": 0.0, "entropy": 0.653888800740242, "epoch": 0.49407114624505927, "frac_reward_zero_std": 0.0, "grad_norm": 1.77239191532135, "learning_rate": 2e-05, "loss": -0.0019287109375, "num_tokens": 7599306.0, "reward": 1.020513916015625, "reward/completion_length_max": 37.2, "reward/completion_length_mean": 30.275, "reward/completion_length_min": 25.0, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.87421875, "reward/mean_r_ndcg": 0.7265625, "reward/mean_r_ndcg_cutoff": 6.4875, "reward/mean_recall": 0.2107421875, "reward/mean_recall_at_10": 0.047802734375, "reward/mean_recall_at_100": 0.2107421875, "reward/mean_retrieval_combined_bounded": 0.02869873046875, "reward/mean_retrieval_combined_raw": 0.0354496568441391, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 122.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.003125, "reward_std": 0.29439025521278384, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.294921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.726085329055786, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.04443097859621048, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.028140258789062, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5323236286640167, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 125, "step_time": 13.751587541389744 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.6375, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.8, "completions/min_terminated_length": 0.0, "entropy": 0.7122428461909294, "epoch": 0.5138339920948617, "frac_reward_zero_std": 0.0, "grad_norm": 1.4968509674072266, "learning_rate": 2e-05, "loss": 0.000927734375, "num_tokens": 7898842.0, "reward": 1.024169921875, "reward/completion_length_max": 39.0, "reward/completion_length_mean": 32.6, "reward/completion_length_min": 26.8, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.009375, "reward/duplicate_ranked_plain_query_rate": 0.009375, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.9890625, "reward/mean_ndcg_at_10": 0.84921875, "reward/mean_r_ndcg": 0.721875, "reward/mean_r_ndcg_cutoff": 6.11875, "reward/mean_recall": 0.2279296875, "reward/mean_recall_at_10": 0.04697265625, "reward/mean_recall_at_100": 0.2279296875, "reward/mean_retrieval_combined_bounded": 0.03355712890625, "reward/mean_retrieval_combined_raw": 0.06411173343658447, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 127.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27766100168228147, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.278125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.5206214427947997, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.024378109723329544, "sampling/raw_vllm_importance_sampling_ratio/min": 1.6010534098585132e-14, "sampling/raw_vllm_sampling_logp_difference/max": 11.999595642089844, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5361509323120117, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 130, "step_time": 13.87040045860922 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.478125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.7116588234901429, "epoch": 0.5335968379446641, "frac_reward_zero_std": 0.0, "grad_norm": 1.5433331727981567, "learning_rate": 2e-05, "loss": 0.0017822265625, "num_tokens": 8198238.0, "reward": 1.0175079345703124, "reward/completion_length_max": 39.0, "reward/completion_length_mean": 32.5, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.83046875, "reward/mean_r_ndcg": 0.67890625, "reward/mean_r_ndcg_cutoff": 5.9625, "reward/mean_recall": 0.2138671875, "reward/mean_recall_at_10": 0.0484375, "reward/mean_recall_at_100": 0.2138671875, "reward/mean_retrieval_combined_bounded": 0.019476318359375, "reward/mean_retrieval_combined_raw": 0.011913469061255455, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 132.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.29369805455207826, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.29375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.597253406047821, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.03286512084305286, "sampling/raw_vllm_importance_sampling_ratio/min": 1.738030059389654e-21, "sampling/raw_vllm_sampling_logp_difference/max": 10.984229278564452, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5033355176448822, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 135, "step_time": 13.938820835598744 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.165625, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.6, "completions/min_terminated_length": 0.0, "entropy": 0.6854275912046432, "epoch": 0.5533596837944664, "frac_reward_zero_std": 0.0, "grad_norm": 1.6360342502593994, "learning_rate": 2e-05, "loss": 0.0016845703125, "num_tokens": 8496146.0, "reward": 1.0207611083984376, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 31.15, "reward/completion_length_min": 26.6, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.8796875, "reward/mean_r_ndcg": 0.74140625, "reward/mean_r_ndcg_cutoff": 6.65, "reward/mean_recall": 0.260546875, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.260546875, "reward/mean_retrieval_combined_bounded": 0.02626953125, "reward/mean_retrieval_combined_raw": 0.03165825456380844, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 137.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.30307801365852355, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.30390625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.9878100633621216, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.0420615665614605, "sampling/raw_vllm_importance_sampling_ratio/min": 2.938112448563852e-17, "sampling/raw_vllm_sampling_logp_difference/max": 11.533337783813476, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5071158826351165, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 140, "step_time": 13.862731645384338 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.5234375, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.8, "completions/min_terminated_length": 0.0, "entropy": 0.6004372745752334, "epoch": 0.5731225296442688, "frac_reward_zero_std": 0.0, "grad_norm": 1.659332036972046, "learning_rate": 2e-05, "loss": -0.0005859375, "num_tokens": 8792528.0, "reward": 1.0218353271484375, "reward/completion_length_max": 36.6, "reward/completion_length_mean": 30.55, "reward/completion_length_min": 25.8, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.975, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.87109375, "reward/mean_r_ndcg": 0.6984375, "reward/mean_r_ndcg_cutoff": 6.05, "reward/mean_recall": 0.2275390625, "reward/mean_recall_at_10": 0.04873046875, "reward/mean_recall_at_100": 0.2275390625, "reward/mean_retrieval_combined_bounded": 0.02193603515625, "reward/mean_retrieval_combined_raw": -1.5925616025924683e-07, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 142.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27633762955665586, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.2765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.9607898831367492, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.043030251376330854, "sampling/raw_vllm_importance_sampling_ratio/min": 3.410978546123278e-14, "sampling/raw_vllm_sampling_logp_difference/max": 12.047430801391602, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5036101579666138, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 145, "step_time": 13.691569352406077 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.6390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.0, "completions/min_terminated_length": 0.0, "entropy": 0.6303751319646835, "epoch": 0.5928853754940712, "frac_reward_zero_std": 0.0, "grad_norm": 1.3859226703643799, "learning_rate": 2e-05, "loss": -0.00038909912109375, "num_tokens": 9087330.0, "reward": 1.0168792724609375, "reward/completion_length_max": 34.6, "reward/completion_length_mean": 29.625, "reward/completion_length_min": 25.0, "reward/completion_length_p90": 32.2, "reward/completion_length_p95": 32.4, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00234375, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01484375, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.9890625, "reward/mean_ndcg_at_10": 0.821875, "reward/mean_r_ndcg": 0.69296875, "reward/mean_r_ndcg_cutoff": 6.28125, "reward/mean_recall": 0.2291015625, "reward/mean_recall_at_10": 0.0474609375, "reward/mean_recall_at_100": 0.2291015625, "reward/mean_retrieval_combined_bounded": 0.0198577880859375, "reward/mean_retrieval_combined_raw": 0.01364363394677639, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 147.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0015625, "reward_std": 0.296415239572525, "rewards/retrieval_reward/mean": 1.01484375, "rewards/retrieval_reward/std": 0.296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.084872007369995, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.026607921719551085, "sampling/raw_vllm_importance_sampling_ratio/min": 2.579603271579644e-15, "sampling/raw_vllm_sampling_logp_difference/max": 10.890928649902344, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5988399147987366, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 150, "step_time": 13.58755170305958 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.02734375, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.0, "completions/min_terminated_length": 0.0, "entropy": 0.5830283313989639, "epoch": 0.5968379446640316, "frac_reward_zero_std": 0.0, "num_tokens": 9145961.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.5871336274673634, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.6412489602443923, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8400702987697716, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9938488576449912, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5480763895198735, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6040134233263232, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9733333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.28838966836504937, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.32720748829117197, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5012048192771085, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8554216867469879, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.5947088794830108, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6025357897165379, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7318086549600634, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9094525486250838, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.566674274500901, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3444041359867099, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17242741991870003, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3172970820084959, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 0.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3452.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.3350167940048948, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.3663875401454093, "retrieval_eval/nq/single_turn/recall_at_10": 0.5468568945538818, "retrieval_eval/nq/single_turn/recall_at_100": 0.810423908845114, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6327219580143522, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6737283442778894, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8286111111111112, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9386666666666668, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.01116943359375, "reward/completion_length_max": 32.0, "reward/completion_length_mean": 28.0, "reward/completion_length_min": 24.0, "reward/completion_length_p90": 31.0, "reward/completion_length_p95": 31.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.01953125, "reward/low_length_query_rate": 0.01953125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0078125, "reward/mean_colbert": 0.95703125, "reward/mean_mrr": 0.98046875, "reward/mean_ndcg_at_10": 0.8203125, "reward/mean_r_ndcg": 0.70703125, "reward/mean_r_ndcg_cutoff": 6.3125, "reward/mean_recall": 0.2138671875, "reward/mean_recall_at_10": 0.044921875, "reward/mean_recall_at_100": 0.2138671875, "reward/mean_retrieval_combined_bounded": 0.04052734375, "reward/mean_retrieval_combined_raw": 0.06689453125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 150.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.01953125, "reward_std": 0.3700309097766876, "rewards/retrieval_reward/mean": 1.0078125, "rewards/retrieval_reward/std": 0.37109375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.0598753690719604, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.01610877923667431, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.286548614501953, "sampling/raw_vllm_sampling_logp_difference/mean": 0.6024929881095886, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 151, "step_time": 13.393256593844853 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 56.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.5615234375, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.0, "completions/min_terminated_length": 0.0, "entropy": 0.4873832240700722, "epoch": 0.6126482213438735, "frac_reward_zero_std": 0.0, "grad_norm": 1.630955696105957, "learning_rate": 2e-05, "loss": 0.004736328125, "num_tokens": 9381128.0, "reward": 1.0026473999023438, "reward/completion_length_max": 56.25, "reward/completion_length_mean": 28.5625, "reward/completion_length_min": 24.0, "reward/completion_length_p90": 30.75, "reward/completion_length_p95": 31.5, "reward/duplicate_query_rate": 0.0126953125, "reward/duplicate_ranked_plain_query_rate": 0.0126953125, "reward/invalid_output_rate": 0.0361328125, "reward/length_out_of_range_rate": 0.0224609375, "reward/low_length_query_rate": 0.0224609375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0009765625, "reward/mean_colbert": 0.939453125, "reward/mean_mrr": 0.9638671875, "reward/mean_ndcg_at_10": 0.8447265625, "reward/mean_r_ndcg": 0.708984375, "reward/mean_r_ndcg_cutoff": 7.09375, "reward/mean_recall": 0.234619140625, "reward/mean_recall_at_10": 0.04638671875, "reward/mean_recall_at_100": 0.234619140625, "reward/mean_retrieval_combined_bounded": 0.05181884765625, "reward/mean_retrieval_combined_raw": 0.098876953125, "reward/output_tag_rate": 0.9990234375, "reward/query_valid_rate": 0.9638671875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9638671875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 152.5, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0224609375, "reward_std": 0.37655678391456604, "rewards/retrieval_reward/mean": 1.0009765625, "rewards/retrieval_reward/std": 0.376953125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.8284814208745956, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.03688519075512886, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.794400453567505, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5588230118155479, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 155, "step_time": 14.579469221498584 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.67421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.2, "completions/min_terminated_length": 0.0, "entropy": 0.5562534026801587, "epoch": 0.6324110671936759, "frac_reward_zero_std": 0.0, "grad_norm": 1.4615861177444458, "learning_rate": 2e-05, "loss": -0.0013916015625, "num_tokens": 9675271.0, "reward": 0.993524169921875, "reward/completion_length_max": 35.6, "reward/completion_length_mean": 28.65, "reward/completion_length_min": 24.2, "reward/completion_length_p90": 31.4, "reward/completion_length_p95": 32.0, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.02265625, "reward/length_out_of_range_rate": 0.0203125, "reward/low_length_query_rate": 0.0203125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.9921875, "reward/mean_colbert": 0.953125, "reward/mean_mrr": 0.97734375, "reward/mean_ndcg_at_10": 0.80625, "reward/mean_r_ndcg": 0.640625, "reward/mean_r_ndcg_cutoff": 5.95625, "reward/mean_recall": 0.2087890625, "reward/mean_recall_at_10": 0.047265625, "reward/mean_recall_at_100": 0.2087890625, "reward/mean_retrieval_combined_bounded": 0.02623443603515625, "reward/mean_retrieval_combined_raw": 0.049269577488303185, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97734375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97734375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 157.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.01953125, "reward_std": 0.33599320650100706, "rewards/retrieval_reward/mean": 0.99296875, "rewards/retrieval_reward/std": 0.33671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.107939124107361, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.04276912864297629, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.587401580810546, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5181473731994629, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 160, "step_time": 13.591869639407378 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.1265625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.0, "completions/min_terminated_length": 0.0, "entropy": 0.5388526603579521, "epoch": 0.6521739130434783, "frac_reward_zero_std": 0.0, "grad_norm": 1.8345028162002563, "learning_rate": 2e-05, "loss": -0.001300048828125, "num_tokens": 9969833.0, "reward": 1.0135406494140624, "reward/completion_length_max": 32.8, "reward/completion_length_mean": 29.125, "reward/completion_length_min": 25.0, "reward/completion_length_p90": 31.2, "reward/completion_length_p95": 31.8, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00546875, "reward/length_out_of_range_rate": 0.003125, "reward/low_length_query_rate": 0.003125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01171875, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.84609375, "reward/mean_r_ndcg": 0.71328125, "reward/mean_r_ndcg_cutoff": 5.08125, "reward/mean_recall": 0.214453125, "reward/mean_recall_at_10": 0.047314453125, "reward/mean_recall_at_100": 0.214453125, "reward/mean_retrieval_combined_bounded": 0.0205780029296875, "reward/mean_retrieval_combined_raw": 0.0318278431892395, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99453125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99453125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 162.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.003125, "reward_std": 0.28717131316661837, "rewards/retrieval_reward/mean": 1.01171875, "rewards/retrieval_reward/std": 0.28828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.939142394065857, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.027212951704859732, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.234678840637207, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5594024002552033, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 165, "step_time": 13.5579807415721 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.19296875, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.6, "completions/min_terminated_length": 0.0, "entropy": 0.5363053813576698, "epoch": 0.6719367588932806, "frac_reward_zero_std": 0.0, "grad_norm": 2.018670082092285, "learning_rate": 2e-05, "loss": -0.00296630859375, "num_tokens": 10265920.0, "reward": 1.0332122802734376, "reward/completion_length_max": 35.8, "reward/completion_length_mean": 30.2, "reward/completion_length_min": 25.6, "reward/completion_length_p90": 32.8, "reward/completion_length_p95": 33.6, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.034375, "reward/mean_colbert": 0.9703125, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.8671875, "reward/mean_r_ndcg": 0.673828125, "reward/mean_r_ndcg_cutoff": 5.14375, "reward/mean_recall": 0.236328125, "reward/mean_recall_at_10": 0.0486328125, "reward/mean_recall_at_100": 0.236328125, "reward/mean_retrieval_combined_bounded": 0.03777999877929687, "reward/mean_retrieval_combined_raw": 0.08896469995379448, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 167.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.29058855175971987, "rewards/retrieval_reward/mean": 1.034375, "rewards/retrieval_reward/std": 0.29140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.0151073813438414, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.02836144305765629, "sampling/raw_vllm_importance_sampling_ratio/min": 4.591114175201652e-15, "sampling/raw_vllm_sampling_logp_difference/max": 9.877704620361328, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5352712988853454, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 170, "step_time": 13.73180029145442 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.49453125, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.8, "completions/min_terminated_length": 0.0, "entropy": 0.5749564185738564, "epoch": 0.691699604743083, "frac_reward_zero_std": 0.0, "grad_norm": 1.3280926942825317, "learning_rate": 2e-05, "loss": -0.00264892578125, "num_tokens": 10563833.0, "reward": 1.013079833984375, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 31.5, "reward/completion_length_min": 26.8, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.871875, "reward/mean_r_ndcg": 0.74453125, "reward/mean_r_ndcg_cutoff": 6.03125, "reward/mean_recall": 0.2203125, "reward/mean_recall_at_10": 0.047607421875, "reward/mean_recall_at_100": 0.2203125, "reward/mean_retrieval_combined_bounded": 0.016168212890625, "reward/mean_retrieval_combined_raw": 0.021143491379916668, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 172.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27593124508857725, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.2765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.1890228390693665, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.024565256759524347, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.96571159362793, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5242570579051972, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 175, "step_time": 13.813146395841613 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.9984375, "completions/max_length": 39.8, "completions/max_terminated_length": 12.2, "completions/mean_length": 33.00859375, "completions/mean_terminated_length": 12.2, "completions/min_length": 27.2, "completions/min_terminated_length": 12.2, "entropy": 0.5927831694483757, "epoch": 0.7114624505928854, "frac_reward_zero_std": 0.0, "grad_norm": 1.7612762451171875, "learning_rate": 2e-05, "loss": -0.00174560546875, "num_tokens": 10863684.0, "reward": 1.024835205078125, "reward/completion_length_max": 39.8, "reward/completion_length_mean": 33.0, "reward/completion_length_min": 27.2, "reward/completion_length_p90": 36.4, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.02421875, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.82734375, "reward/mean_r_ndcg": 0.671875, "reward/mean_r_ndcg_cutoff": 6.3, "reward/mean_recall": 0.2345703125, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.2345703125, "reward/mean_retrieval_combined_bounded": 0.033740234375, "reward/mean_retrieval_combined_raw": 0.1196533203125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 177.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.2967938244342804, "rewards/retrieval_reward/mean": 1.02421875, "rewards/retrieval_reward/std": 0.296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.387757349014282, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.04577840678393841, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.475551223754882, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4894444584846497, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 180, "step_time": 13.913306762394495 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.528125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.2, "completions/min_terminated_length": 0.0, "entropy": 0.5754422694444656, "epoch": 0.7312252964426877, "frac_reward_zero_std": 0.0, "grad_norm": 1.8151166439056396, "learning_rate": 2e-05, "loss": 0.0016845703125, "num_tokens": 11164264.0, "reward": 1.0175689697265624, "reward/completion_length_max": 40.0, "reward/completion_length_mean": 33.5, "reward/completion_length_min": 29.2, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 37.2, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.890625, "reward/mean_r_ndcg": 0.7234375, "reward/mean_r_ndcg_cutoff": 6.15625, "reward/mean_recall": 0.2291015625, "reward/mean_recall_at_10": 0.047998046875, "reward/mean_recall_at_100": 0.2291015625, "reward/mean_retrieval_combined_bounded": 0.02254638671875, "reward/mean_retrieval_combined_raw": 0.02357899770140648, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 182.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.2805400490760803, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.2806640625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.0744075775146484, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.04513409361243248, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.45382480621338, "sampling/raw_vllm_sampling_logp_difference/mean": 0.44222607016563414, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 185, "step_time": 13.969954394374508 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.88828125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.8, "completions/min_terminated_length": 0.0, "entropy": 0.5801244854927063, "epoch": 0.7509881422924901, "frac_reward_zero_std": 0.0, "grad_norm": 1.6337522268295288, "learning_rate": 2e-05, "loss": 0.001446533203125, "num_tokens": 11463737.0, "reward": 1.005596923828125, "reward/completion_length_max": 40.8, "reward/completion_length_mean": 32.9, "reward/completion_length_min": 27.8, "reward/completion_length_p90": 36.2, "reward/completion_length_p95": 37.4, "reward/duplicate_query_rate": 0.0109375, "reward/duplicate_ranked_plain_query_rate": 0.0109375, "reward/invalid_output_rate": 0.0203125, "reward/length_out_of_range_rate": 0.009375, "reward/low_length_query_rate": 0.009375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00546875, "reward/mean_colbert": 0.95546875, "reward/mean_mrr": 0.978125, "reward/mean_ndcg_at_10": 0.809375, "reward/mean_r_ndcg": 0.67734375, "reward/mean_r_ndcg_cutoff": 5.8875, "reward/mean_recall": 0.248828125, "reward/mean_recall_at_10": 0.0470703125, "reward/mean_recall_at_100": 0.248828125, "reward/mean_retrieval_combined_bounded": 0.0305908203125, "reward/mean_retrieval_combined_raw": 0.05576166324317455, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9796875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9796875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 187.0, "reward/search_failure_rate": 0.0015625, "reward/too_long_query_rate": 0.009375, "reward/too_short_query_rate": 0.0, "reward_std": 0.32188771963119506, "rewards/retrieval_reward/mean": 1.00546875, "rewards/retrieval_reward/std": 0.323046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.1374054670333864, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.05426798239350319, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.21323127746582, "sampling/raw_vllm_sampling_logp_difference/mean": 0.449777227640152, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 190, "step_time": 13.980974368401803 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.5609375, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.8, "completions/min_terminated_length": 0.0, "entropy": 0.5276537723839283, "epoch": 0.7707509881422925, "frac_reward_zero_std": 0.0, "grad_norm": 1.6878979206085205, "learning_rate": 2e-05, "loss": -9.46044921875e-05, "num_tokens": 11761287.0, "reward": 1.0152679443359376, "reward/completion_length_max": 38.8, "reward/completion_length_mean": 31.575, "reward/completion_length_min": 26.8, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.003125, "reward/low_length_query_rate": 0.003125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.8640625, "reward/mean_r_ndcg": 0.74375, "reward/mean_r_ndcg_cutoff": 6.6, "reward/mean_recall": 0.2375, "reward/mean_recall_at_10": 0.04873046875, "reward/mean_recall_at_100": 0.2375, "reward/mean_retrieval_combined_bounded": 0.023138427734375, "reward/mean_retrieval_combined_raw": 0.029199268855154514, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 192.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.00234375, "reward_std": 0.2919617176055908, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.2921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.6740398406982422, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.02780722863972187, "sampling/raw_vllm_importance_sampling_ratio/min": 4.891432062812601e-21, "sampling/raw_vllm_sampling_logp_difference/max": 10.717801666259765, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5294947981834411, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 195, "step_time": 13.89907629264053 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.115625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.8, "completions/min_terminated_length": 0.0, "entropy": 0.5168608792126179, "epoch": 0.7905138339920948, "frac_reward_zero_std": 0.0, "grad_norm": 1.9322564601898193, "learning_rate": 2e-05, "loss": 0.0004150390625, "num_tokens": 12057467.0, "reward": 1.0126922607421875, "reward/completion_length_max": 36.4, "reward/completion_length_mean": 30.1, "reward/completion_length_min": 25.8, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 33.6, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01171875, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.8703125, "reward/mean_r_ndcg": 0.7125, "reward/mean_r_ndcg_cutoff": 6.15625, "reward/mean_recall": 0.233984375, "reward/mean_recall_at_10": 0.0490234375, "reward/mean_recall_at_100": 0.233984375, "reward/mean_retrieval_combined_bounded": 0.01593017578125, "reward/mean_retrieval_combined_raw": 0.016991253942251205, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 197.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.29164178371429444, "rewards/retrieval_reward/mean": 1.01171875, "rewards/retrieval_reward/std": 0.291796875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.934180212020874, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.0411152970045805, "sampling/raw_vllm_importance_sampling_ratio/min": 2.0454302382161454e-15, "sampling/raw_vllm_sampling_logp_difference/max": 10.421612358093261, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5426645994186401, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 200, "step_time": 13.814181490580086 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.99609375, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.0, "completions/min_terminated_length": 0.0, "entropy": 0.50517837703228, "epoch": 0.7944664031620553, "frac_reward_zero_std": 0.0, "num_tokens": 12116314.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6316265138724321, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.6861153472543826, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8778558875219684, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9938488576449912, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5607170362687622, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6189578808483671, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.82, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3053658537863314, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3520998718191228, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5445783132530121, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8578313253012049, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6096444051011818, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6283310430394093, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7695517803792451, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9031983568075117, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5894364027759922, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3603280321317758, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17966253912853564, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.329992544854575, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 0.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3452.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.36953666060749085, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4051236319511583, "retrieval_eval/nq/single_turn/recall_at_10": 0.5966348010814987, "retrieval_eval/nq/single_turn/recall_at_100": 0.8544563538045578, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6564237444272297, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6958632931497357, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8456666666666667, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9516666666666667, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 0.9947509765625, "reward/completion_length_max": 33.0, "reward/completion_length_mean": 29.0, "reward/completion_length_min": 26.0, "reward/completion_length_p90": 31.0, "reward/completion_length_p95": 32.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99609375, "reward/mean_colbert": 0.9765625, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.8984375, "reward/mean_r_ndcg": 0.73828125, "reward/mean_r_ndcg_cutoff": 4.875, "reward/mean_recall": 0.20703125, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.20703125, "reward/mean_retrieval_combined_bounded": -0.005157470703125, "reward/mean_retrieval_combined_raw": -3.8743019104003906e-07, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 200.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28355175256729126, "rewards/retrieval_reward/mean": 0.99609375, "rewards/retrieval_reward/std": 0.283203125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.590519905090332, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.029788384214043617, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.09739875793457, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5936515927314758, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 201, "step_time": 13.584979882929474 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.9931640625, "completions/mean_terminated_length": 0.0, "completions/min_length": 23.75, "completions/min_terminated_length": 0.0, "entropy": 0.48670730646699667, "epoch": 0.8102766798418972, "frac_reward_zero_std": 0.0, "grad_norm": 2.7089483737945557, "learning_rate": 2e-05, "loss": -0.0041015625, "num_tokens": 12352275.0, "reward": 0.9892578125, "reward/completion_length_max": 35.25, "reward/completion_length_mean": 29.0, "reward/completion_length_min": 23.75, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 32.5, "reward/duplicate_query_rate": 0.0048828125, "reward/duplicate_ranked_plain_query_rate": 0.0048828125, "reward/invalid_output_rate": 0.03515625, "reward/length_out_of_range_rate": 0.0302734375, "reward/low_length_query_rate": 0.0302734375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.9892578125, "reward/mean_colbert": 0.94140625, "reward/mean_mrr": 0.96484375, "reward/mean_ndcg_at_10": 0.8447265625, "reward/mean_r_ndcg": 0.6689453125, "reward/mean_r_ndcg_cutoff": 5.7109375, "reward/mean_recall": 0.226318359375, "reward/mean_recall_at_10": 0.04718017578125, "reward/mean_recall_at_100": 0.226318359375, "reward/mean_retrieval_combined_bounded": 0.0395050048828125, "reward/mean_retrieval_combined_raw": 0.08155066519975662, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.96484375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96484375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 202.5, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0302734375, "reward_std": 0.3742222934961319, "rewards/retrieval_reward/mean": 0.9892578125, "rewards/retrieval_reward/std": 0.37451171875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.8178243041038513, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.033666857983917, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.344449520111084, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5532257556915283, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 205, "step_time": 13.686264249758096 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.7421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.6, "completions/min_terminated_length": 0.0, "entropy": 0.532669098675251, "epoch": 0.8300395256916996, "frac_reward_zero_std": 0.0, "grad_norm": 2.0743677616119385, "learning_rate": 2e-05, "loss": -0.00126953125, "num_tokens": 12648137.0, "reward": 1.006768798828125, "reward/completion_length_max": 35.2, "reward/completion_length_mean": 29.75, "reward/completion_length_min": 24.6, "reward/completion_length_p90": 32.2, "reward/completion_length_p95": 32.8, "reward/duplicate_query_rate": 0.00859375, "reward/duplicate_ranked_plain_query_rate": 0.00859375, "reward/invalid_output_rate": 0.0234375, "reward/length_out_of_range_rate": 0.01484375, "reward/low_length_query_rate": 0.01484375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00703125, "reward/mean_colbert": 0.95234375, "reward/mean_mrr": 0.9765625, "reward/mean_ndcg_at_10": 0.85546875, "reward/mean_r_ndcg": 0.68515625, "reward/mean_r_ndcg_cutoff": 6.3375, "reward/mean_recall": 0.2126953125, "reward/mean_recall_at_10": 0.04775390625, "reward/mean_recall_at_100": 0.2126953125, "reward/mean_retrieval_combined_bounded": 0.0376220703125, "reward/mean_retrieval_combined_raw": 0.05432431995868683, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 207.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.01484375, "reward_std": 0.3420539855957031, "rewards/retrieval_reward/mean": 1.00546875, "rewards/retrieval_reward/std": 0.342578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.327376961708069, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.04418524336069822, "sampling/raw_vllm_importance_sampling_ratio/min": 9.01281106186753e-18, "sampling/raw_vllm_sampling_logp_difference/max": 12.38114948272705, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5628700971603393, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 210, "step_time": 13.689513054757844 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.653125, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.8, "completions/min_terminated_length": 0.0, "entropy": 0.5269144892692565, "epoch": 0.849802371541502, "frac_reward_zero_std": 0.0, "grad_norm": 2.0692331790924072, "learning_rate": 2e-05, "loss": 0.00150146484375, "num_tokens": 12944717.0, "reward": 1.0071136474609375, "reward/completion_length_max": 36.8, "reward/completion_length_mean": 30.625, "reward/completion_length_min": 25.8, "reward/completion_length_p90": 33.4, "reward/completion_length_p95": 34.2, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.00546875, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00703125, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.98203125, "reward/mean_ndcg_at_10": 0.84609375, "reward/mean_r_ndcg": 0.703125, "reward/mean_r_ndcg_cutoff": 6.5125, "reward/mean_recall": 0.25, "reward/mean_recall_at_10": 0.0482421875, "reward/mean_recall_at_100": 0.25, "reward/mean_retrieval_combined_bounded": 0.01282958984375, "reward/mean_retrieval_combined_raw": -0.011522850394248963, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99453125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99453125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 212.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.2911245107650757, "rewards/retrieval_reward/mean": 1.00703125, "rewards/retrieval_reward/std": 0.29140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.1451518058776855, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.02161165000870824, "sampling/raw_vllm_importance_sampling_ratio/min": 5.310379889186638e-17, "sampling/raw_vllm_sampling_logp_difference/max": 11.55309009552002, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5823675036430359, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 215, "step_time": 13.765210951620247 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 55.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.30390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.2, "completions/min_terminated_length": 0.0, "entropy": 0.5482429467141628, "epoch": 0.8695652173913043, "frac_reward_zero_std": 0.0, "grad_norm": 2.2811927795410156, "learning_rate": 2e-05, "loss": 0.013671875, "num_tokens": 13241010.0, "reward": 1.01749267578125, "reward/completion_length_max": 55.4, "reward/completion_length_mean": 30.3, "reward/completion_length_min": 25.2, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.90078125, "reward/mean_r_ndcg": 0.7734375, "reward/mean_r_ndcg_cutoff": 7.19375, "reward/mean_recall": 0.24765625, "reward/mean_recall_at_10": 0.049072265625, "reward/mean_recall_at_100": 0.24765625, "reward/mean_retrieval_combined_bounded": 0.02808837890625, "reward/mean_retrieval_combined_raw": 0.037224340438842776, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 217.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.003125, "reward_std": 0.3062672197818756, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.306640625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.9777841329574586, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.02187586631625891, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 13.84235725402832, "sampling/raw_vllm_sampling_logp_difference/mean": 0.6170165300369262, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 220, "step_time": 14.600238503410946 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.70625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.4, "completions/min_terminated_length": 0.0, "entropy": 0.528595182299614, "epoch": 0.8893280632411067, "frac_reward_zero_std": 0.0, "grad_norm": 2.5912458896636963, "learning_rate": 2e-05, "loss": -0.00059814453125, "num_tokens": 13536794.0, "reward": 1.024798583984375, "reward/completion_length_max": 36.2, "reward/completion_length_mean": 29.675, "reward/completion_length_min": 25.4, "reward/completion_length_p90": 32.2, "reward/completion_length_p95": 32.8, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.90546875, "reward/mean_r_ndcg": 0.7609375, "reward/mean_r_ndcg_cutoff": 6.89375, "reward/mean_recall": 0.2236328125, "reward/mean_recall_at_10": 0.049267578125, "reward/mean_recall_at_100": 0.2236328125, "reward/mean_retrieval_combined_bounded": 0.02640228271484375, "reward/mean_retrieval_combined_raw": 0.012793394178152085, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 222.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2724658191204071, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.2732421875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.0469781279563906, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.028410109132528304, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 14.635145568847657, "sampling/raw_vllm_sampling_logp_difference/mean": 0.6161634206771851, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 225, "step_time": 13.732248078018893 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.6, "completions/min_terminated_length": 0.0, "entropy": 0.5238786920905113, "epoch": 0.9090909090909091, "frac_reward_zero_std": 0.0, "grad_norm": 2.5134835243225098, "learning_rate": 2e-05, "loss": -0.0035888671875, "num_tokens": 13830998.0, "reward": 1.021380615234375, "reward/completion_length_max": 34.4, "reward/completion_length_mean": 28.925, "reward/completion_length_min": 24.6, "reward/completion_length_p90": 31.2, "reward/completion_length_p95": 31.8, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.96640625, "reward/mean_mrr": 0.990625, "reward/mean_ndcg_at_10": 0.865625, "reward/mean_r_ndcg": 0.7203125, "reward/mean_r_ndcg_cutoff": 6.1375, "reward/mean_recall": 0.218359375, "reward/mean_recall_at_10": 0.047021484375, "reward/mean_recall_at_100": 0.218359375, "reward/mean_retrieval_combined_bounded": 0.0320068359375, "reward/mean_retrieval_combined_raw": 0.04279781766235828, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 227.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00234375, "reward_std": 0.28898478150367735, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.289453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.033023351430893, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.029467657022178174, "sampling/raw_vllm_importance_sampling_ratio/min": 1.1039493903980228e-15, "sampling/raw_vllm_sampling_logp_difference/max": 16.086458206176758, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5943145275115966, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 230, "step_time": 13.615374798362609 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 43.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.0390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.4, "completions/min_terminated_length": 0.0, "entropy": 0.5004955567419529, "epoch": 0.9288537549407114, "frac_reward_zero_std": 0.0, "grad_norm": 3.2270305156707764, "learning_rate": 2e-05, "loss": 0.008984375, "num_tokens": 14126856.0, "reward": 1.01444091796875, "reward/completion_length_max": 43.2, "reward/completion_length_mean": 30.05, "reward/completion_length_min": 25.4, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 33.2, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.02109375, "reward/length_out_of_range_rate": 0.0140625, "reward/low_length_query_rate": 0.0140625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01328125, "reward/mean_colbert": 0.9546875, "reward/mean_mrr": 0.97578125, "reward/mean_ndcg_at_10": 0.83203125, "reward/mean_r_ndcg": 0.683203125, "reward/mean_r_ndcg_cutoff": 5.88125, "reward/mean_recall": 0.2453125, "reward/mean_recall_at_10": 0.047412109375, "reward/mean_recall_at_100": 0.2453125, "reward/mean_retrieval_combined_bounded": 0.04251708984375, "reward/mean_retrieval_combined_raw": 0.07207213640213013, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 232.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00859375, "reward/too_short_query_rate": 0.00546875, "reward_std": 0.3300721764564514, "rewards/retrieval_reward/mean": 1.01328125, "rewards/retrieval_reward/std": 0.33046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4623754978179933, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.02958734631538391, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 14.677625274658203, "sampling/raw_vllm_sampling_logp_difference/mean": 0.6198894619941712, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 235, "step_time": 14.096904742228798 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.98984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.2, "completions/min_terminated_length": 0.0, "entropy": 0.49109645560383797, "epoch": 0.9486166007905138, "frac_reward_zero_std": 0.0, "grad_norm": 2.804048776626587, "learning_rate": 2e-05, "loss": -0.00272216796875, "num_tokens": 14422651.0, "reward": 0.999591064453125, "reward/completion_length_max": 35.4, "reward/completion_length_mean": 29.975, "reward/completion_length_min": 25.2, "reward/completion_length_p90": 32.4, "reward/completion_length_p95": 33.2, "reward/duplicate_query_rate": 0.03984375, "reward/duplicate_ranked_plain_query_rate": 0.03984375, "reward/invalid_output_rate": 0.0421875, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.9984375, "reward/mean_colbert": 0.934375, "reward/mean_mrr": 0.9578125, "reward/mean_ndcg_at_10": 0.815625, "reward/mean_r_ndcg": 0.6421875, "reward/mean_r_ndcg_cutoff": 5.35, "reward/mean_recall": 0.2203125, "reward/mean_recall_at_10": 0.046142578125, "reward/mean_recall_at_100": 0.2203125, "reward/mean_retrieval_combined_bounded": 0.042647171020507815, "reward/mean_retrieval_combined_raw": 0.09638671875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9578125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9578125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 237.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00234375, "reward_std": 0.33127491772174833, "rewards/retrieval_reward/mean": 1.0, "rewards/retrieval_reward/std": 0.33203125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.9838883519172668, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.030472220107913017, "sampling/raw_vllm_importance_sampling_ratio/min": 2.768738722138461e-14, "sampling/raw_vllm_sampling_logp_difference/max": 16.755540466308595, "sampling/raw_vllm_sampling_logp_difference/mean": 0.6107715249061585, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 240, "step_time": 13.609083599061705 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.559375, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.6, "completions/min_terminated_length": 0.0, "entropy": 0.49999830946326257, "epoch": 0.9683794466403162, "frac_reward_zero_std": 0.0, "grad_norm": 2.648963689804077, "learning_rate": 2e-05, "loss": -0.0031982421875, "num_tokens": 14719495.0, "reward": 1.0187744140625, "reward/completion_length_max": 37.2, "reward/completion_length_mean": 30.575, "reward/completion_length_min": 25.6, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 34.2, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.85, "reward/mean_r_ndcg": 0.6875, "reward/mean_r_ndcg_cutoff": 5.8125, "reward/mean_recall": 0.2263671875, "reward/mean_recall_at_10": 0.048193359375, "reward/mean_recall_at_100": 0.2263671875, "reward/mean_retrieval_combined_bounded": 0.0257171630859375, "reward/mean_retrieval_combined_raw": 0.03632615059614182, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 242.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.28608805537223814, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.28671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.7280651569366454, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.02701529152691364, "sampling/raw_vllm_importance_sampling_ratio/min": 1.8217933344347513e-18, "sampling/raw_vllm_sampling_logp_difference/max": 15.308475875854493, "sampling/raw_vllm_sampling_logp_difference/mean": 0.6036610245704651, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 245, "step_time": 13.79269829619443 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.78359375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.4, "completions/min_terminated_length": 0.0, "entropy": 0.485552953928709, "epoch": 0.9881422924901185, "frac_reward_zero_std": 0.0, "grad_norm": 2.2559640407562256, "learning_rate": 2e-05, "loss": 0.00159912109375, "num_tokens": 15017586.0, "reward": 1.0095123291015624, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 31.75, "reward/completion_length_min": 27.4, "reward/completion_length_p90": 34.8, "reward/completion_length_p95": 35.4, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01015625, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.88125, "reward/mean_r_ndcg": 0.72734375, "reward/mean_r_ndcg_cutoff": 5.64375, "reward/mean_recall": 0.229296875, "reward/mean_recall_at_10": 0.048193359375, "reward/mean_recall_at_100": 0.229296875, "reward/mean_retrieval_combined_bounded": 0.01156005859375, "reward/mean_retrieval_combined_raw": -0.06289427392184735, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 247.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2807962954044342, "rewards/retrieval_reward/mean": 1.01015625, "rewards/retrieval_reward/std": 0.280859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.6161168932914733, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.022688647359609605, "sampling/raw_vllm_importance_sampling_ratio/min": 9.031195187332232e-24, "sampling/raw_vllm_sampling_logp_difference/max": 10.945592880249023, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5478050529956817, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 250, "step_time": 13.834904071036727 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.5609516948461533, "epoch": 0.9920948616600791, "frac_reward_zero_std": 0.0, "num_tokens": 15077270.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.5947487381931307, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.6484784319676282, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.844463971880492, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9912126537785588, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5546705935506677, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6075657776064273, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.98, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.30905148877216976, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3498108963232287, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5228915662650603, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8481927710843373, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6155821001317132, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6320909120707318, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7717162825437472, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9160253185781355, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5634379038204459, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3512911252387664, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17341461408450226, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3225414821105539, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 0.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3452.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.3747137192788983, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.41013411753525947, "retrieval_eval/nq/single_turn/recall_at_10": 0.6012215140981073, "retrieval_eval/nq/single_turn/recall_at_100": 0.8583912707609115, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6541111247689538, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6911132816509006, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8377777777777778, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9420000000000001, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.097808837890625, "reward/completion_length_max": 41.0, "reward/completion_length_mean": 32.25, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.09375, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.89453125, "reward/mean_r_ndcg": 0.79296875, "reward/mean_r_ndcg_cutoff": 6.96875, "reward/mean_recall": 0.24609375, "reward/mean_recall_at_10": 0.049560546875, "reward/mean_recall_at_100": 0.24609375, "reward/mean_retrieval_combined_bounded": 0.10302734375, "reward/mean_retrieval_combined_raw": 0.404296875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 250.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00390625, "reward/too_short_query_rate": 0.0, "reward_std": 0.29411405324935913, "rewards/retrieval_reward/mean": 1.1015625, "rewards/retrieval_reward/std": 0.294921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5749518871307373, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.032767727971076965, "sampling/raw_vllm_importance_sampling_ratio/min": 3.589392690267172e-18, "sampling/raw_vllm_sampling_logp_difference/max": 14.564221382141113, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5295862555503845, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 251, "step_time": 14.028444688010495 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 60.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.392578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.5316839898005128, "epoch": 1.007905138339921, "frac_reward_zero_std": 0.0, "grad_norm": 1.9882451295852661, "learning_rate": 2e-05, "loss": 0.01875, "num_tokens": 15316552.0, "reward": 0.9928550720214844, "reward/completion_length_max": 60.0, "reward/completion_length_mean": 32.40625, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.0498046875, "reward/duplicate_ranked_plain_query_rate": 0.0498046875, "reward/invalid_output_rate": 0.0556640625, "reward/length_out_of_range_rate": 0.001953125, "reward/low_length_query_rate": 0.001953125, "reward/malformed_query_rate": 0.0009765625, "reward/malformed_ranked_plain_query_rate": 0.0009765625, "reward/mean": 0.9931640625, "reward/mean_colbert": 0.9208984375, "reward/mean_mrr": 0.9443359375, "reward/mean_ndcg_at_10": 0.79296875, "reward/mean_r_ndcg": 0.580078125, "reward/mean_r_ndcg_cutoff": 5.359375, "reward/mean_recall": 0.213623046875, "reward/mean_recall_at_10": 0.04510498046875, "reward/mean_recall_at_100": 0.213623046875, "reward/mean_retrieval_combined_bounded": 0.056243896484375, "reward/mean_retrieval_combined_raw": 0.12656331062316895, "reward/output_tag_rate": 0.9970703125, "reward/query_valid_rate": 0.9443359375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9443359375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 252.5, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.001953125, "reward/too_short_query_rate": 0.0, "reward_std": 0.364862322807312, "rewards/retrieval_reward/mean": 0.9931640625, "rewards/retrieval_reward/std": 0.365234375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.2191730737686157, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.033654564060270786, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 15.655502080917358, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5036610662937164, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 255, "step_time": 14.871357235024334 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.24375, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.8, "completions/min_terminated_length": 0.0, "entropy": 0.5601696446537971, "epoch": 1.0276679841897234, "frac_reward_zero_std": 0.0, "grad_norm": 1.8731507062911987, "learning_rate": 2e-05, "loss": -0.0023193359375, "num_tokens": 15615072.0, "reward": 1.0194366455078125, "reward/completion_length_max": 39.2, "reward/completion_length_mean": 32.275, "reward/completion_length_min": 26.8, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.01015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.9640625, "reward/mean_mrr": 0.98984375, "reward/mean_ndcg_at_10": 0.8375, "reward/mean_r_ndcg": 0.70703125, "reward/mean_r_ndcg_cutoff": 6.36875, "reward/mean_recall": 0.2416015625, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.2416015625, "reward/mean_retrieval_combined_bounded": 0.0296905517578125, "reward/mean_retrieval_combined_raw": 0.044580194912850854, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 257.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2854061722755432, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.2859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.6662120342254638, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.033248480036854745, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.554874229431153, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5029088973999023, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 260, "step_time": 13.888652429147623 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.38828125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.5904149778187275, "epoch": 1.0474308300395256, "frac_reward_zero_std": 0.0, "grad_norm": 2.200042724609375, "learning_rate": 2e-05, "loss": 0.00242919921875, "num_tokens": 15914225.0, "reward": 1.0305450439453125, "reward/completion_length_max": 41.0, "reward/completion_length_mean": 32.35, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 35.4, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0296875, "reward/mean_colbert": 0.96640625, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.903125, "reward/mean_r_ndcg": 0.7671875, "reward/mean_r_ndcg_cutoff": 6.9625, "reward/mean_recall": 0.222265625, "reward/mean_recall_at_10": 0.04853515625, "reward/mean_recall_at_100": 0.222265625, "reward/mean_retrieval_combined_bounded": 0.03948974609375, "reward/mean_retrieval_combined_raw": 0.05085288286209107, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 262.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.2799722611904144, "rewards/retrieval_reward/mean": 1.0296875, "rewards/retrieval_reward/std": 0.28046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.8171713471412658, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.033407197892665864, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.939900970458984, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5059265911579132, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 265, "step_time": 13.994231197820046 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.05703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.8, "completions/min_terminated_length": 0.0, "entropy": 0.5702902741730214, "epoch": 1.0671936758893281, "frac_reward_zero_std": 0.0, "grad_norm": 2.156906843185425, "learning_rate": 2e-05, "loss": 0.0002593994140625, "num_tokens": 16212474.0, "reward": 1.011279296875, "reward/completion_length_max": 38.8, "reward/completion_length_mean": 32.075, "reward/completion_length_min": 26.8, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0109375, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.87265625, "reward/mean_r_ndcg": 0.7234375, "reward/mean_r_ndcg_cutoff": 5.715625, "reward/mean_recall": 0.248828125, "reward/mean_recall_at_10": 0.048388671875, "reward/mean_recall_at_100": 0.248828125, "reward/mean_retrieval_combined_bounded": 0.014788818359375, "reward/mean_retrieval_combined_raw": 0.024438700079917906, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 266.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.274748432636261, "rewards/retrieval_reward/mean": 1.0109375, "rewards/retrieval_reward/std": 0.274609375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.9732833623886108, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.0466778289526701, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.770243644714355, "sampling/raw_vllm_sampling_logp_difference/mean": 0.458418071269989, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 270, "step_time": 13.873984370997642 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.1078125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.5830901071429253, "epoch": 1.0869565217391304, "frac_reward_zero_std": 0.0, "grad_norm": 2.1212007999420166, "learning_rate": 2e-05, "loss": 0.00272216796875, "num_tokens": 16510852.0, "reward": 1.0276763916015625, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 32.125, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.9703125, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.86171875, "reward/mean_r_ndcg": 0.753125, "reward/mean_r_ndcg_cutoff": 6.00625, "reward/mean_recall": 0.230859375, "reward/mean_recall_at_10": 0.04873046875, "reward/mean_recall_at_100": 0.230859375, "reward/mean_retrieval_combined_bounded": 0.03233642578125, "reward/mean_retrieval_combined_raw": 0.10041497144848108, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 272.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.29458208084106446, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.294921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.0340314388275145, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.054036623984575274, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.931054306030274, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4286763250827789, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 275, "step_time": 13.853013765020297 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.87265625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.8, "completions/min_terminated_length": 0.0, "entropy": 0.5636268466711044, "epoch": 1.1067193675889329, "frac_reward_zero_std": 0.0, "grad_norm": 2.23795747756958, "learning_rate": 2e-05, "loss": -0.00018768310546875, "num_tokens": 16809441.0, "reward": 1.0203582763671875, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 31.85, "reward/completion_length_min": 25.8, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 36.2, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.02109375, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.875, "reward/mean_r_ndcg": 0.72109375, "reward/mean_r_ndcg_cutoff": 5.975, "reward/mean_recall": 0.2197265625, "reward/mean_recall_at_10": 0.048974609375, "reward/mean_recall_at_100": 0.2197265625, "reward/mean_retrieval_combined_bounded": 0.022048187255859376, "reward/mean_retrieval_combined_raw": 0.04779718481004238, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 277.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2798071324825287, "rewards/retrieval_reward/mean": 1.02109375, "rewards/retrieval_reward/std": 0.2796875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.8788763761520386, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.04587954953312874, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.940070533752442, "sampling/raw_vllm_sampling_logp_difference/mean": 0.45141568779945374, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 280, "step_time": 13.880619804619347 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.4640625, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.0, "completions/min_terminated_length": 0.0, "entropy": 0.5455805592238903, "epoch": 1.1264822134387351, "frac_reward_zero_std": 0.0, "grad_norm": 2.005303382873535, "learning_rate": 2e-05, "loss": 0.00184326171875, "num_tokens": 17107347.0, "reward": 1.0183807373046876, "reward/completion_length_max": 39.4, "reward/completion_length_mean": 31.475, "reward/completion_length_min": 26.0, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.01640625, "reward/length_out_of_range_rate": 0.00859375, "reward/low_length_query_rate": 0.00859375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.95859375, "reward/mean_mrr": 0.98359375, "reward/mean_ndcg_at_10": 0.86015625, "reward/mean_r_ndcg": 0.70625, "reward/mean_r_ndcg_cutoff": 6.31875, "reward/mean_recall": 0.223046875, "reward/mean_recall_at_10": 0.0478515625, "reward/mean_recall_at_100": 0.223046875, "reward/mean_retrieval_combined_bounded": 0.039068603515625, "reward/mean_retrieval_combined_raw": 0.063427734375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98359375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98359375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 282.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00859375, "reward/too_short_query_rate": 0.0, "reward_std": 0.31911468505859375, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.31953125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.422758460044861, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.05235209539532661, "sampling/raw_vllm_importance_sampling_ratio/min": 3.4896690165363636e-16, "sampling/raw_vllm_sampling_logp_difference/max": 11.833518791198731, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4746661365032196, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 285, "step_time": 13.879121234442573 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.4, "completions/min_terminated_length": 0.0, "entropy": 0.5844276860356331, "epoch": 1.1462450592885376, "frac_reward_zero_std": 0.0, "grad_norm": 2.170607805252075, "learning_rate": 2e-05, "loss": -0.001434326171875, "num_tokens": 17404391.0, "reward": 1.0196136474609374, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 31.15, "reward/completion_length_min": 25.4, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.0109375, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.9890625, "reward/mean_ndcg_at_10": 0.88359375, "reward/mean_r_ndcg": 0.7984375, "reward/mean_r_ndcg_cutoff": 6.49375, "reward/mean_recall": 0.2119140625, "reward/mean_recall_at_10": 0.04873046875, "reward/mean_recall_at_100": 0.2119140625, "reward/mean_retrieval_combined_bounded": 0.030877685546875, "reward/mean_retrieval_combined_raw": 0.04527778625488281, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 286.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.2874439835548401, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.287890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.823149871826172, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.08511440455913544, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.784199142456055, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4344915091991425, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 290, "step_time": 13.792394321365283 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 42.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.6578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.592048679292202, "epoch": 1.1660079051383399, "frac_reward_zero_std": 0.0, "grad_norm": 2.2064743041992188, "learning_rate": 2e-05, "loss": 0.006591796875, "num_tokens": 17703505.0, "reward": 0.9899017333984375, "reward/completion_length_max": 42.6, "reward/completion_length_mean": 32.65, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 36.8, "reward/completion_length_p95": 38.8, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.03203125, "reward/length_out_of_range_rate": 0.0265625, "reward/low_length_query_rate": 0.0265625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.98984375, "reward/mean_colbert": 0.94453125, "reward/mean_mrr": 0.9671875, "reward/mean_ndcg_at_10": 0.83046875, "reward/mean_r_ndcg": 0.740625, "reward/mean_r_ndcg_cutoff": 6.11875, "reward/mean_recall": 0.22421875, "reward/mean_recall_at_10": 0.04638671875, "reward/mean_recall_at_100": 0.22421875, "reward/mean_retrieval_combined_bounded": 0.035205078125, "reward/mean_retrieval_combined_raw": 0.07036127746105195, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.96796875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96796875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 292.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0265625, "reward/too_short_query_rate": 0.0, "reward_std": 0.36333602070808413, "rewards/retrieval_reward/mean": 0.98984375, "rewards/retrieval_reward/std": 0.36328125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.8922523736953736, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.03714421465992927, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.678192710876464, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4873351871967316, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 295, "step_time": 14.064151894568932 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.11484375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.2, "completions/min_terminated_length": 0.0, "entropy": 0.6127538025379181, "epoch": 1.1857707509881423, "frac_reward_zero_std": 0.0, "grad_norm": 2.055816411972046, "learning_rate": 2e-05, "loss": 0.005224609375, "num_tokens": 18001828.0, "reward": 1.0180694580078125, "reward/completion_length_max": 40.8, "reward/completion_length_mean": 32.1, "reward/completion_length_min": 27.2, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.0109375, "reward/duplicate_ranked_plain_query_rate": 0.0109375, "reward/invalid_output_rate": 0.02421875, "reward/length_out_of_range_rate": 0.01328125, "reward/low_length_query_rate": 0.01328125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.95234375, "reward/mean_mrr": 0.97578125, "reward/mean_ndcg_at_10": 0.88359375, "reward/mean_r_ndcg": 0.76015625, "reward/mean_r_ndcg_cutoff": 6.28125, "reward/mean_recall": 0.2294921875, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.2294921875, "reward/mean_retrieval_combined_bounded": 0.04893798828125, "reward/mean_retrieval_combined_raw": 0.09155399203300477, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97578125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97578125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 297.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01328125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3281419098377228, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.328515625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.275597405433655, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.040864995867013934, "sampling/raw_vllm_importance_sampling_ratio/min": 1.2290721292755983e-21, "sampling/raw_vllm_sampling_logp_difference/max": 12.95766544342041, "sampling/raw_vllm_sampling_logp_difference/mean": 0.48277589678764343, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 300, "step_time": 13.903596493788063 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.359375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.5799987018108368, "epoch": 1.1897233201581028, "frac_reward_zero_std": 0.0, "num_tokens": 18061632.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6251605974578002, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.6816164385491238, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8787346221441125, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9947275922671354, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6015188654798506, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6521290256908933, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8333333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.31793706824869244, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.36257544420030047, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5469879518072289, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8433734939759037, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6105753258268937, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6226421187522243, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7558068105603317, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9219776995305163, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5835584156650596, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3655256571481995, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.18098434056690627, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3276702681293454, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 2.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3450.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.3925029984524622, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.43220613450525097, "retrieval_eval/nq/single_turn/recall_at_10": 0.634975845410628, "retrieval_eval/nq/single_turn/recall_at_100": 0.8751449275362319, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6669615236827625, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7015303934391498, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8422222222222222, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9426666666666667, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0000762939453125, "reward/completion_length_max": 39.0, "reward/completion_length_mean": 32.25, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.91015625, "reward/mean_r_ndcg": 0.703125, "reward/mean_r_ndcg_cutoff": 6.75, "reward/mean_recall": 0.2080078125, "reward/mean_recall_at_10": 0.04931640625, "reward/mean_recall_at_100": 0.2080078125, "reward/mean_retrieval_combined_bounded": 0.0002040863037109375, "reward/mean_retrieval_combined_raw": -1.52587890625e-05, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 300.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28704822063446045, "rewards/retrieval_reward/mean": 1.0, "rewards/retrieval_reward/std": 0.287109375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.2903553247451782, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.02837245538830757, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.893935203552246, "sampling/raw_vllm_sampling_logp_difference/mean": 0.46811434626579285, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 301, "step_time": 13.894421749049798 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.75, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.4697265625, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.5, "completions/min_terminated_length": 0.0, "entropy": 0.5792466383427382, "epoch": 1.2055335968379446, "frac_reward_zero_std": 0.0, "grad_norm": 2.010012149810791, "learning_rate": 2e-05, "loss": 0.0007476806640625, "num_tokens": 18300833.0, "reward": 1.0241661071777344, "reward/completion_length_max": 38.75, "reward/completion_length_mean": 32.4375, "reward/completion_length_min": 27.5, "reward/completion_length_p90": 35.25, "reward/completion_length_p95": 36.25, "reward/duplicate_query_rate": 0.001953125, "reward/duplicate_ranked_plain_query_rate": 0.001953125, "reward/invalid_output_rate": 0.001953125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.9736328125, "reward/mean_mrr": 0.998046875, "reward/mean_ndcg_at_10": 0.919921875, "reward/mean_r_ndcg": 0.837890625, "reward/mean_r_ndcg_cutoff": 6.5234375, "reward/mean_recall": 0.233154296875, "reward/mean_recall_at_10": 0.0484619140625, "reward/mean_recall_at_100": 0.233154296875, "reward/mean_retrieval_combined_bounded": 0.0262298583984375, "reward/mean_retrieval_combined_raw": 0.015122465789318085, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.998046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.998046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 302.5, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2730809673666954, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.2734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.3721636235713959, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.019866410410031676, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.267175197601318, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5534918904304504, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 305, "step_time": 13.918777206999948 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.7921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.0, "completions/min_terminated_length": 0.0, "entropy": 0.6184296831488609, "epoch": 1.225296442687747, "frac_reward_zero_std": 0.0, "grad_norm": 2.388949155807495, "learning_rate": 2e-05, "loss": 0.0034912109375, "num_tokens": 18600023.0, "reward": 0.99866943359375, "reward/completion_length_max": 41.2, "reward/completion_length_mean": 32.775, "reward/completion_length_min": 26.0, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 37.8, "reward/duplicate_query_rate": 0.02890625, "reward/duplicate_ranked_plain_query_rate": 0.02890625, "reward/invalid_output_rate": 0.04375, "reward/length_out_of_range_rate": 0.01484375, "reward/low_length_query_rate": 0.01484375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99921875, "reward/mean_colbert": 0.9328125, "reward/mean_mrr": 0.95546875, "reward/mean_ndcg_at_10": 0.8265625, "reward/mean_r_ndcg": 0.70390625, "reward/mean_r_ndcg_cutoff": 6.4, "reward/mean_recall": 0.221484375, "reward/mean_recall_at_10": 0.04521484375, "reward/mean_recall_at_100": 0.221484375, "reward/mean_retrieval_combined_bounded": 0.0498291015625, "reward/mean_retrieval_combined_raw": 0.09990544319152832, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.95625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.95625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 306.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01484375, "reward/too_short_query_rate": 0.0, "reward_std": 0.3602580904960632, "rewards/retrieval_reward/mean": 0.99921875, "rewards/retrieval_reward/std": 0.361328125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.9332022190093994, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.03136813379824162, "sampling/raw_vllm_importance_sampling_ratio/min": 1.9268656674009642e-20, "sampling/raw_vllm_sampling_logp_difference/max": 10.285103797912598, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5423884689807892, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 310, "step_time": 13.982030447816943 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.809375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.5689908273518085, "epoch": 1.2450592885375493, "frac_reward_zero_std": 0.0, "grad_norm": 3.12953782081604, "learning_rate": 2e-05, "loss": 0.00069580078125, "num_tokens": 18899779.0, "reward": 1.033709716796875, "reward/completion_length_max": 38.8, "reward/completion_length_mean": 32.8, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.00859375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0015625, "reward/malformed_ranked_plain_query_rate": 0.0015625, "reward/mean": 1.0328125, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.99140625, "reward/mean_ndcg_at_10": 0.8515625, "reward/mean_r_ndcg": 0.68984375, "reward/mean_r_ndcg_cutoff": 5.40625, "reward/mean_recall": 0.23125, "reward/mean_recall_at_10": 0.04853515625, "reward/mean_recall_at_100": 0.23125, "reward/mean_retrieval_combined_bounded": 0.04384765625, "reward/mean_retrieval_combined_raw": 0.11201154799200594, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99140625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99140625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 312.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.29983721375465394, "rewards/retrieval_reward/mean": 1.0328125, "rewards/retrieval_reward/std": 0.3, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.6018753111362458, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.01893481407314539, "sampling/raw_vllm_importance_sampling_ratio/min": 3.3823221376837235e-20, "sampling/raw_vllm_sampling_logp_difference/max": 12.936099624633789, "sampling/raw_vllm_sampling_logp_difference/mean": 0.600628137588501, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 315, "step_time": 13.907223314035217 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 57.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.5359375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.5440883748233318, "epoch": 1.2648221343873518, "frac_reward_zero_std": 0.0, "grad_norm": 2.8879454135894775, "learning_rate": 2e-05, "loss": 0.012890625, "num_tokens": 19198993.0, "reward": 1.0023162841796875, "reward/completion_length_max": 57.0, "reward/completion_length_mean": 32.6, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.6, "reward/duplicate_query_rate": 0.01484375, "reward/duplicate_ranked_plain_query_rate": 0.01484375, "reward/invalid_output_rate": 0.03125, "reward/length_out_of_range_rate": 0.015625, "reward/low_length_query_rate": 0.015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00234375, "reward/mean_colbert": 0.9453125, "reward/mean_mrr": 0.96875, "reward/mean_ndcg_at_10": 0.8359375, "reward/mean_r_ndcg": 0.6671875, "reward/mean_r_ndcg_cutoff": 5.2375, "reward/mean_recall": 0.2234375, "reward/mean_recall_at_10": 0.047265625, "reward/mean_recall_at_100": 0.2234375, "reward/mean_retrieval_combined_bounded": 0.0429931640625, "reward/mean_retrieval_combined_raw": 0.080419921875, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.96875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 317.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.353591126203537, "rewards/retrieval_reward/mean": 1.00234375, "rewards/retrieval_reward/std": 0.35390625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.5733442187309266, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.021162048913538456, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.597406959533691, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5995583772659302, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 320, "step_time": 14.675500641018152 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.6890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.4, "completions/min_terminated_length": 0.0, "entropy": 0.5578580550849438, "epoch": 1.2845849802371543, "frac_reward_zero_std": 0.0, "grad_norm": 2.166337251663208, "learning_rate": 2e-05, "loss": 0.0005340576171875, "num_tokens": 19496963.0, "reward": 1.0184539794921874, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 31.675, "reward/completion_length_min": 27.4, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.02578125, "reward/duplicate_ranked_plain_query_rate": 0.02578125, "reward/invalid_output_rate": 0.02578125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.9515625, "reward/mean_mrr": 0.9703125, "reward/mean_ndcg_at_10": 0.86875, "reward/mean_r_ndcg": 0.73828125, "reward/mean_r_ndcg_cutoff": 6.475, "reward/mean_recall": 0.2416015625, "reward/mean_recall_at_10": 0.04755859375, "reward/mean_recall_at_100": 0.2416015625, "reward/mean_retrieval_combined_bounded": 0.0444091796875, "reward/mean_retrieval_combined_raw": 0.092822265625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97421875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97421875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 322.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.31085665225982667, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.31171875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.3307680487632751, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.012299024779349566, "sampling/raw_vllm_importance_sampling_ratio/min": 7.408563230004505e-19, "sampling/raw_vllm_sampling_logp_difference/max": 14.394227981567383, "sampling/raw_vllm_sampling_logp_difference/mean": 0.6460584402084351, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 325, "step_time": 13.779137749399524 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.52578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.8, "completions/min_terminated_length": 0.0, "entropy": 0.5318762481212616, "epoch": 1.3043478260869565, "frac_reward_zero_std": 0.0, "grad_norm": 2.316500663757324, "learning_rate": 2e-05, "loss": 0.00033721923828125, "num_tokens": 19793220.0, "reward": 1.0040863037109375, "reward/completion_length_max": 36.6, "reward/completion_length_mean": 30.5, "reward/completion_length_min": 24.8, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.01875, "reward/duplicate_ranked_plain_query_rate": 0.01875, "reward/invalid_output_rate": 0.025, "reward/length_out_of_range_rate": 0.00625, "reward/low_length_query_rate": 0.00625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0046875, "reward/mean_colbert": 0.95078125, "reward/mean_mrr": 0.97421875, "reward/mean_ndcg_at_10": 0.85078125, "reward/mean_r_ndcg": 0.68125, "reward/mean_r_ndcg_cutoff": 5.85625, "reward/mean_recall": 0.2259765625, "reward/mean_recall_at_10": 0.046826171875, "reward/mean_recall_at_100": 0.2259765625, "reward/mean_retrieval_combined_bounded": 0.0322021484375, "reward/mean_retrieval_combined_raw": 0.06767578125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.975, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.975, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 326.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.00546875, "reward_std": 0.3239101588726044, "rewards/retrieval_reward/mean": 1.0046875, "rewards/retrieval_reward/std": 0.32421875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.9814400672912598, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.02946970146149397, "sampling/raw_vllm_importance_sampling_ratio/min": 7.421760431388694e-21, "sampling/raw_vllm_sampling_logp_difference/max": 12.415609741210938, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5865894913673401, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 330, "step_time": 13.690163258998655 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.940625, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.0, "completions/min_terminated_length": 0.0, "entropy": 0.5084358096122742, "epoch": 1.3241106719367588, "frac_reward_zero_std": 0.0, "grad_norm": 3.4853076934814453, "learning_rate": 2e-05, "loss": -0.0062255859375, "num_tokens": 20087320.0, "reward": 1.0156097412109375, "reward/completion_length_max": 35.4, "reward/completion_length_mean": 28.925, "reward/completion_length_min": 24.0, "reward/completion_length_p90": 31.4, "reward/completion_length_p95": 32.4, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.0234375, "reward/length_out_of_range_rate": 0.01640625, "reward/low_length_query_rate": 0.01640625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01796875, "reward/mean_colbert": 0.953125, "reward/mean_mrr": 0.9765625, "reward/mean_ndcg_at_10": 0.89765625, "reward/mean_r_ndcg": 0.815625, "reward/mean_r_ndcg_cutoff": 7.03125, "reward/mean_recall": 0.221875, "reward/mean_recall_at_10": 0.048193359375, "reward/mean_recall_at_100": 0.221875, "reward/mean_retrieval_combined_bounded": 0.047265625, "reward/mean_retrieval_combined_raw": 0.090673828125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 332.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.01640625, "reward_std": 0.3391316533088684, "rewards/retrieval_reward/mean": 1.01796875, "rewards/retrieval_reward/std": 0.33984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.192097747325897, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.023015696741640567, "sampling/raw_vllm_importance_sampling_ratio/min": 4.758782072048012e-19, "sampling/raw_vllm_sampling_logp_difference/max": 15.032955169677734, "sampling/raw_vllm_sampling_logp_difference/mean": 0.6271532058715821, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 335, "step_time": 13.617898496834096 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.8375, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.2, "completions/min_terminated_length": 0.0, "entropy": 0.4860316962003708, "epoch": 1.3438735177865613, "frac_reward_zero_std": 0.0, "grad_norm": 2.477968692779541, "learning_rate": 2e-05, "loss": 0.00098876953125, "num_tokens": 20382888.0, "reward": 1.0293548583984375, "reward/completion_length_max": 37.0, "reward/completion_length_mean": 29.825, "reward/completion_length_min": 24.2, "reward/completion_length_p90": 32.8, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.028125, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.85546875, "reward/mean_r_ndcg": 0.70546875, "reward/mean_r_ndcg_cutoff": 5.4875, "reward/mean_recall": 0.215234375, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.215234375, "reward/mean_retrieval_combined_bounded": 0.03223876953125, "reward/mean_retrieval_combined_raw": 0.12270615398883819, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 337.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.29886448979377744, "rewards/retrieval_reward/mean": 1.028125, "rewards/retrieval_reward/std": 0.299609375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.045431971549988, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.042826133966445926, "sampling/raw_vllm_importance_sampling_ratio/min": 4.6526725614712875e-14, "sampling/raw_vllm_sampling_logp_difference/max": 13.743692588806152, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5458344638347625, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 340, "step_time": 13.762495087587741 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.68984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.8, "completions/min_terminated_length": 0.0, "entropy": 0.491594535112381, "epoch": 1.3636363636363638, "frac_reward_zero_std": 0.05, "grad_norm": 1.9430526494979858, "learning_rate": 2e-05, "loss": -5.035400390625e-05, "num_tokens": 20678619.0, "reward": 1.035186767578125, "reward/completion_length_max": 37.6, "reward/completion_length_mean": 29.7, "reward/completion_length_min": 24.8, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.015625, "reward/duplicate_ranked_plain_query_rate": 0.015625, "reward/invalid_output_rate": 0.0234375, "reward/length_out_of_range_rate": 0.0078125, "reward/low_length_query_rate": 0.0078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03359375, "reward/mean_colbert": 0.95234375, "reward/mean_mrr": 0.97578125, "reward/mean_ndcg_at_10": 0.84921875, "reward/mean_r_ndcg": 0.715625, "reward/mean_r_ndcg_cutoff": 5.81875, "reward/mean_recall": 0.2431640625, "reward/mean_recall_at_10": 0.047216796875, "reward/mean_recall_at_100": 0.2431640625, "reward/mean_retrieval_combined_bounded": 0.062451171875, "reward/mean_retrieval_combined_raw": 0.2916015625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 342.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.00703125, "reward_std": 0.3308578968048096, "rewards/retrieval_reward/mean": 1.03359375, "rewards/retrieval_reward/std": 0.331640625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.9224321722984314, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.04389756396412849, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.835920333862305, "sampling/raw_vllm_sampling_logp_difference/mean": 0.50414337515831, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 345, "step_time": 13.751529642415699 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.77578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.2, "completions/min_terminated_length": 0.0, "entropy": 0.4448386698961258, "epoch": 1.383399209486166, "frac_reward_zero_std": 0.025, "grad_norm": 1.950251579284668, "learning_rate": 2e-05, "loss": -0.00118408203125, "num_tokens": 20973692.0, "reward": 0.9962188720703125, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 29.775, "reward/completion_length_min": 25.2, "reward/completion_length_p90": 33.4, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.01171875, "reward/duplicate_ranked_plain_query_rate": 0.01171875, "reward/invalid_output_rate": 0.0296875, "reward/length_out_of_range_rate": 0.0171875, "reward/low_length_query_rate": 0.0171875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.996875, "reward/mean_colbert": 0.946875, "reward/mean_mrr": 0.9703125, "reward/mean_ndcg_at_10": 0.84609375, "reward/mean_r_ndcg": 0.69453125, "reward/mean_r_ndcg_cutoff": 5.4, "reward/mean_recall": 0.231640625, "reward/mean_recall_at_10": 0.04765625, "reward/mean_recall_at_100": 0.231640625, "reward/mean_retrieval_combined_bounded": 0.0361572265625, "reward/mean_retrieval_combined_raw": 0.06845664475113153, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.9703125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9703125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 346.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01640625, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.350160014629364, "rewards/retrieval_reward/mean": 0.996875, "rewards/retrieval_reward/std": 0.35, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3130399227142333, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.05850733295083046, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 14.735056114196777, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4669513523578644, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 350, "step_time": 13.770200003415813 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 42.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.7421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.0, "completions/min_terminated_length": 0.0, "entropy": 0.45554403215646744, "epoch": 1.3873517786561265, "frac_reward_zero_std": 0.0, "num_tokens": 21032954.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6450598289279137, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.6998854982623071, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8892794376098418, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9947275922671354, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5914583533680047, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6465307370168528, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.84, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.322685271991219, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3688219897717473, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5566265060240964, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8481927710843373, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6114852922320761, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6297856574291928, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7676361197487959, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9166121730382293, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5881995073411391, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.36653818510008823, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.18008956918237584, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.32972783168157677, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 2.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3450.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.3827638948109258, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4186337502712787, "retrieval_eval/nq/single_turn/recall_at_10": 0.6154347826086957, "retrieval_eval/nq/single_turn/recall_at_100": 0.8734299516908213, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6423831925604901, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6817480031689804, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8343333333333334, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9470000000000001, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0416107177734375, "reward/completion_length_max": 42.0, "reward/completion_length_mean": 30.75, "reward/completion_length_min": 25.0, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0390625, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.83203125, "reward/mean_r_ndcg": 0.6484375, "reward/mean_r_ndcg_cutoff": 5.4375, "reward/mean_recall": 0.26171875, "reward/mean_recall_at_10": 0.049560546875, "reward/mean_recall_at_100": 0.26171875, "reward/mean_retrieval_combined_bounded": 0.051513671875, "reward/mean_retrieval_combined_raw": 0.0625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 350.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00390625, "reward/too_short_query_rate": 0.0, "reward_std": 0.28660252690315247, "rewards/retrieval_reward/mean": 1.0390625, "rewards/retrieval_reward/std": 0.287109375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8846054077148438, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.06741388887166977, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.759603500366211, "sampling/raw_vllm_sampling_logp_difference/mean": 0.41112929582595825, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 351, "step_time": 13.957687645044643 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 60.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.70703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.75, "completions/min_terminated_length": 0.0, "entropy": 0.4855507640168071, "epoch": 1.4031620553359683, "frac_reward_zero_std": 0.0, "grad_norm": 2.5361547470092773, "learning_rate": 2e-05, "loss": 0.0140625, "num_tokens": 21270286.0, "reward": 1.010498046875, "reward/completion_length_max": 60.0, "reward/completion_length_mean": 30.6875, "reward/completion_length_min": 26.75, "reward/completion_length_p90": 33.25, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.0244140625, "reward/duplicate_ranked_plain_query_rate": 0.0244140625, "reward/invalid_output_rate": 0.0263671875, "reward/length_out_of_range_rate": 0.0009765625, "reward/low_length_query_rate": 0.0009765625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01171875, "reward/mean_colbert": 0.9501953125, "reward/mean_mrr": 0.9736328125, "reward/mean_ndcg_at_10": 0.876953125, "reward/mean_r_ndcg": 0.70703125, "reward/mean_r_ndcg_cutoff": 6.296875, "reward/mean_recall": 0.202392578125, "reward/mean_recall_at_10": 0.04718017578125, "reward/mean_recall_at_100": 0.202392578125, "reward/mean_retrieval_combined_bounded": 0.039642333984375, "reward/mean_retrieval_combined_raw": 0.09246826171875, "reward/output_tag_rate": 0.9990234375, "reward/query_valid_rate": 0.9736328125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9736328125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 352.5, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0009765625, "reward_std": 0.3265664651989937, "rewards/retrieval_reward/mean": 1.01171875, "rewards/retrieval_reward/std": 0.3271484375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.1897035241127014, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.05428154952824116, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.82950210571289, "sampling/raw_vllm_sampling_logp_difference/mean": 0.466180182993412, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 355, "step_time": 14.824231869977666 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.66796875, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.8, "completions/min_terminated_length": 0.0, "entropy": 0.4978739835321903, "epoch": 1.4229249011857708, "frac_reward_zero_std": 0.0, "grad_norm": 1.8164358139038086, "learning_rate": 2e-05, "loss": -0.004150390625, "num_tokens": 21565701.0, "reward": 0.99022216796875, "reward/completion_length_max": 35.0, "reward/completion_length_mean": 29.65, "reward/completion_length_min": 24.8, "reward/completion_length_p90": 32.2, "reward/completion_length_p95": 32.6, "reward/duplicate_query_rate": 0.0375, "reward/duplicate_ranked_plain_query_rate": 0.0375, "reward/invalid_output_rate": 0.05, "reward/length_out_of_range_rate": 0.0109375, "reward/low_length_query_rate": 0.0109375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.98984375, "reward/mean_colbert": 0.92578125, "reward/mean_mrr": 0.94921875, "reward/mean_ndcg_at_10": 0.83828125, "reward/mean_r_ndcg": 0.72734375, "reward/mean_r_ndcg_cutoff": 6.71875, "reward/mean_recall": 0.2216796875, "reward/mean_recall_at_10": 0.04677734375, "reward/mean_recall_at_100": 0.2216796875, "reward/mean_retrieval_combined_bounded": 0.04862060546875, "reward/mean_retrieval_combined_raw": 0.10634765893919393, "reward/output_tag_rate": 0.9984375, "reward/query_valid_rate": 0.95, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.95, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 357.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0109375, "reward_std": 0.3708180248737335, "rewards/retrieval_reward/mean": 0.98984375, "rewards/retrieval_reward/std": 0.372265625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3790236949920653, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.06339490786194801, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.698736953735352, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4589205801486969, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 360, "step_time": 13.625853363820351 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.80859375, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.8, "completions/min_terminated_length": 0.0, "entropy": 0.4328146010637283, "epoch": 1.4426877470355732, "frac_reward_zero_std": 0.0, "grad_norm": 2.1367416381835938, "learning_rate": 2e-05, "loss": -0.00380859375, "num_tokens": 21860080.0, "reward": 1.022802734375, "reward/completion_length_max": 34.2, "reward/completion_length_mean": 28.8, "reward/completion_length_min": 24.8, "reward/completion_length_p90": 31.0, "reward/completion_length_p95": 31.6, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.01171875, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.96328125, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.8421875, "reward/mean_r_ndcg": 0.66328125, "reward/mean_r_ndcg_cutoff": 5.5, "reward/mean_recall": 0.2412109375, "reward/mean_recall_at_10": 0.0478515625, "reward/mean_recall_at_100": 0.2412109375, "reward/mean_retrieval_combined_bounded": 0.038818359375, "reward/mean_retrieval_combined_raw": 0.075, "reward/output_tag_rate": 0.9984375, "reward/query_valid_rate": 0.98828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 362.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00234375, "reward_std": 0.29999869465827944, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.30078125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4008824825286865, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.084999780356884, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.946355819702148, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4347060561180115, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 365, "step_time": 13.614339770807419 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.59375, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.4, "completions/min_terminated_length": 0.0, "entropy": 0.4063492015004158, "epoch": 1.4624505928853755, "frac_reward_zero_std": 0.0, "grad_norm": 2.8289873600006104, "learning_rate": 2e-05, "loss": -0.00157470703125, "num_tokens": 22154472.0, "reward": 1.0150848388671876, "reward/completion_length_max": 34.8, "reward/completion_length_mean": 28.575, "reward/completion_length_min": 24.4, "reward/completion_length_p90": 31.2, "reward/completion_length_p95": 32.6, "reward/duplicate_query_rate": 0.0109375, "reward/duplicate_ranked_plain_query_rate": 0.0109375, "reward/invalid_output_rate": 0.02109375, "reward/length_out_of_range_rate": 0.01015625, "reward/low_length_query_rate": 0.01015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.95546875, "reward/mean_mrr": 0.97890625, "reward/mean_ndcg_at_10": 0.875, "reward/mean_r_ndcg": 0.70703125, "reward/mean_r_ndcg_cutoff": 5.83125, "reward/mean_recall": 0.2302734375, "reward/mean_recall_at_10": 0.0484375, "reward/mean_recall_at_100": 0.2302734375, "reward/mean_retrieval_combined_bounded": 0.0412841796875, "reward/mean_retrieval_combined_raw": 0.07172864302992821, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 366.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.01015625, "reward_std": 0.30936195254325866, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.309765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.312758946418762, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.08314978256821633, "sampling/raw_vllm_importance_sampling_ratio/min": 5.680920833661457e-20, "sampling/raw_vllm_sampling_logp_difference/max": 12.759362411499023, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4470307409763336, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 370, "step_time": 13.58506761342287 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.30703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.2, "completions/min_terminated_length": 0.0, "entropy": 0.4337416052818298, "epoch": 1.4822134387351777, "frac_reward_zero_std": 0.0, "grad_norm": 1.865606427192688, "learning_rate": 2e-05, "loss": -0.001025390625, "num_tokens": 22449233.0, "reward": 1.0275726318359375, "reward/completion_length_max": 35.2, "reward/completion_length_mean": 29.3, "reward/completion_length_min": 25.2, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.00859375, "reward/duplicate_ranked_plain_query_rate": 0.00859375, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.01015625, "reward/low_length_query_rate": 0.01015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.02890625, "reward/mean_colbert": 0.95546875, "reward/mean_mrr": 0.9796875, "reward/mean_ndcg_at_10": 0.84609375, "reward/mean_r_ndcg": 0.66484375, "reward/mean_r_ndcg_cutoff": 5.50625, "reward/mean_recall": 0.225, "reward/mean_recall_at_10": 0.047119140625, "reward/mean_recall_at_100": 0.225, "reward/mean_retrieval_combined_bounded": 0.0536376953125, "reward/mean_retrieval_combined_raw": 0.14898648262023925, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 372.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.01015625, "reward_std": 0.3380797028541565, "rewards/retrieval_reward/mean": 1.02890625, "rewards/retrieval_reward/std": 0.3390625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3160737991333007, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.052761290967464444, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 13.156759643554688, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4477992594242096, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 375, "step_time": 13.637368769769092 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 58.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.92265625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.5088281653821468, "epoch": 1.5019762845849802, "frac_reward_zero_std": 0.025, "grad_norm": 1.6603703498840332, "learning_rate": 2e-05, "loss": 0.01162109375, "num_tokens": 22748974.0, "reward": 0.9864654541015625, "reward/completion_length_max": 58.0, "reward/completion_length_mean": 32.9, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 35.4, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.06640625, "reward/duplicate_ranked_plain_query_rate": 0.06640625, "reward/invalid_output_rate": 0.06953125, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.00078125, "reward/malformed_ranked_plain_query_rate": 0.00078125, "reward/mean": 0.9859375, "reward/mean_colbert": 0.90703125, "reward/mean_mrr": 0.92734375, "reward/mean_ndcg_at_10": 0.82578125, "reward/mean_r_ndcg": 0.73046875, "reward/mean_r_ndcg_cutoff": 6.1625, "reward/mean_recall": 0.226171875, "reward/mean_recall_at_10": 0.04521484375, "reward/mean_recall_at_100": 0.226171875, "reward/mean_retrieval_combined_bounded": 0.05914306640625, "reward/mean_retrieval_combined_raw": 0.12768434286117553, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.93046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.93046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 377.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3802169978618622, "rewards/retrieval_reward/mean": 0.9859375, "rewards/retrieval_reward/std": 0.38125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.415101981163025, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.06837627068161964, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.449483108520507, "sampling/raw_vllm_sampling_logp_difference/mean": 0.38969714641571046, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 380, "step_time": 14.64825782260159 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.5609375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.57611895352602, "epoch": 1.5217391304347827, "frac_reward_zero_std": 0.0, "grad_norm": 2.2049548625946045, "learning_rate": 2e-05, "loss": 0.00260009765625, "num_tokens": 23049628.0, "reward": 1.020233154296875, "reward/completion_length_max": 39.8, "reward/completion_length_mean": 33.55, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 36.4, "reward/completion_length_p95": 37.4, "reward/duplicate_query_rate": 0.0171875, "reward/duplicate_ranked_plain_query_rate": 0.0171875, "reward/invalid_output_rate": 0.025, "reward/length_out_of_range_rate": 0.0078125, "reward/low_length_query_rate": 0.0078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.95078125, "reward/mean_mrr": 0.975, "reward/mean_ndcg_at_10": 0.88203125, "reward/mean_r_ndcg": 0.74140625, "reward/mean_r_ndcg_cutoff": 7.06875, "reward/mean_recall": 0.241015625, "reward/mean_recall_at_10": 0.04794921875, "reward/mean_recall_at_100": 0.241015625, "reward/mean_retrieval_combined_bounded": 0.04920654296875, "reward/mean_retrieval_combined_raw": 0.112646484375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.975, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.975, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 382.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3225721836090088, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.32265625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3495247602462768, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.052134844660758975, "sampling/raw_vllm_importance_sampling_ratio/min": 4.628753841806263e-15, "sampling/raw_vllm_sampling_logp_difference/max": 11.27017707824707, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4454879641532898, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 385, "step_time": 13.954786228854209 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.834375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.5032039992511272, "epoch": 1.541501976284585, "frac_reward_zero_std": 0.0, "grad_norm": 2.8540761470794678, "learning_rate": 2e-05, "loss": 0.00302734375, "num_tokens": 23349480.0, "reward": 1.0081573486328126, "reward/completion_length_max": 40.8, "reward/completion_length_mean": 32.875, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 37.6, "reward/duplicate_query_rate": 0.015625, "reward/duplicate_ranked_plain_query_rate": 0.015625, "reward/invalid_output_rate": 0.0296875, "reward/length_out_of_range_rate": 0.0140625, "reward/low_length_query_rate": 0.0140625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0078125, "reward/mean_colbert": 0.94453125, "reward/mean_mrr": 0.9671875, "reward/mean_ndcg_at_10": 0.834375, "reward/mean_r_ndcg": 0.64765625, "reward/mean_r_ndcg_cutoff": 5.56875, "reward/mean_recall": 0.2080078125, "reward/mean_recall_at_10": 0.046044921875, "reward/mean_recall_at_100": 0.2080078125, "reward/mean_retrieval_combined_bounded": 0.0449951171875, "reward/mean_retrieval_combined_raw": 0.08253045082092285, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9703125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9703125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 386.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0140625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3338545858860016, "rewards/retrieval_reward/mean": 1.0078125, "rewards/retrieval_reward/std": 0.334765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.514134740829468, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.043660924583673474, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.588961982727051, "sampling/raw_vllm_sampling_logp_difference/mean": 0.47217877507209777, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 390, "step_time": 13.92999607496895 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 56.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.4015625, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.6, "completions/min_terminated_length": 0.0, "entropy": 0.505259457975626, "epoch": 1.5612648221343872, "frac_reward_zero_std": 0.0, "grad_norm": 2.0314769744873047, "learning_rate": 2e-05, "loss": 0.016015625, "num_tokens": 23646890.0, "reward": 1.0245513916015625, "reward/completion_length_max": 56.8, "reward/completion_length_mean": 31.375, "reward/completion_length_min": 26.6, "reward/completion_length_p90": 33.8, "reward/completion_length_p95": 34.6, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.01171875, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.9640625, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.8453125, "reward/mean_r_ndcg": 0.6984375, "reward/mean_r_ndcg_cutoff": 5.94375, "reward/mean_recall": 0.2443359375, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.2443359375, "reward/mean_retrieval_combined_bounded": 0.0382568359375, "reward/mean_retrieval_combined_raw": 0.048699939250946046, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.98828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 392.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2985355854034424, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.298046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.498359441757202, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.05219558626413345, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.127248573303223, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4735095977783203, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 395, "step_time": 14.637802611431107 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.26328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.4, "completions/min_terminated_length": 0.0, "entropy": 0.4675962343811989, "epoch": 1.5810276679841897, "frac_reward_zero_std": 0.0, "grad_norm": 1.9883699417114258, "learning_rate": 2e-05, "loss": -0.00145263671875, "num_tokens": 23943291.0, "reward": 1.0069732666015625, "reward/completion_length_max": 35.2, "reward/completion_length_mean": 30.275, "reward/completion_length_min": 26.4, "reward/completion_length_p90": 32.8, "reward/completion_length_p95": 33.4, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00625, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.99765625, "reward/mean_ndcg_at_10": 0.86953125, "reward/mean_r_ndcg": 0.73046875, "reward/mean_r_ndcg_cutoff": 5.709375, "reward/mean_recall": 0.226171875, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.226171875, "reward/mean_retrieval_combined_bounded": 0.0093658447265625, "reward/mean_retrieval_combined_raw": 0.0019057709723711014, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 397.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2763070404529572, "rewards/retrieval_reward/mean": 1.00625, "rewards/retrieval_reward/std": 0.276171875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.045949935913086, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.06542090252041817, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.381940269470215, "sampling/raw_vllm_sampling_logp_difference/mean": 0.408799535036087, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 400, "step_time": 13.63396882439265 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.01171875, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.0, "completions/min_terminated_length": 0.0, "entropy": 0.4845997095108032, "epoch": 1.5849802371541502, "frac_reward_zero_std": 0.0, "num_tokens": 24002334.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6388047799542881, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.6935066064142387, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8840070298769771, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9947275922671354, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5861432230519655, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6344612869039684, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8066666666666666, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3270406166498002, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3681801145483329, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5421686746987951, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8530120481927711, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6126473529793276, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6275519840826999, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7654354155234437, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9289361167002012, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5850650144523597, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.37058828727803195, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.18422430780231383, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.32932400086137426, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 2.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3450.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.3828376239900511, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.42072334172347625, "retrieval_eval/nq/single_turn/recall_at_10": 0.6235265700483091, "retrieval_eval/nq/single_turn/recall_at_100": 0.8745893719806763, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6517944586744691, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.691367173496702, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8427777777777778, "retrieval_eval/scifact/single_turn/recall_at_100": 0.938, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0279541015625, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 30.0, "reward/completion_length_min": 26.0, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.0234375, "reward/duplicate_ranked_plain_query_rate": 0.0234375, "reward/invalid_output_rate": 0.0234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.953125, "reward/mean_mrr": 0.9765625, "reward/mean_ndcg_at_10": 0.7890625, "reward/mean_r_ndcg": 0.5703125, "reward/mean_r_ndcg_cutoff": 4.125, "reward/mean_recall": 0.2236328125, "reward/mean_recall_at_10": 0.0458984375, "reward/mean_recall_at_100": 0.2236328125, "reward/mean_retrieval_combined_bounded": 0.05126953125, "reward/mean_retrieval_combined_raw": 0.0986328125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 400.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2994016706943512, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.298828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.513791561126709, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1137489527463913, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.646817207336426, "sampling/raw_vllm_sampling_logp_difference/mean": 0.36825528740882874, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 401, "step_time": 13.753672106016893 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 59.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.3701171875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.5, "completions/min_terminated_length": 0.0, "entropy": 0.4839798389002681, "epoch": 1.6007905138339922, "frac_reward_zero_std": 0.0, "grad_norm": 1.820568561553955, "learning_rate": 2e-05, "loss": 0.0125, "num_tokens": 24240761.0, "reward": 1.0205001831054688, "reward/completion_length_max": 59.25, "reward/completion_length_mean": 31.40625, "reward/completion_length_min": 27.5, "reward/completion_length_p90": 33.75, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.0029296875, "reward/duplicate_ranked_plain_query_rate": 0.0029296875, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021484375, "reward/mean_colbert": 0.9716796875, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.8857421875, "reward/mean_r_ndcg": 0.7177734375, "reward/mean_r_ndcg_cutoff": 6.296875, "reward/mean_recall": 0.22802734375, "reward/mean_recall_at_10": 0.04901123046875, "reward/mean_recall_at_100": 0.22802734375, "reward/mean_retrieval_combined_bounded": 0.0264739990234375, "reward/mean_retrieval_combined_raw": 0.026184496469795704, "reward/output_tag_rate": 0.9990234375, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 402.5, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.30909232050180435, "rewards/retrieval_reward/mean": 1.021484375, "rewards/retrieval_reward/std": 0.31005859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3563196659088135, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.0917256185784936, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.925522327423096, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3743916302919388, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 405, "step_time": 14.85152873073821 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.81328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.2, "completions/min_terminated_length": 0.0, "entropy": 0.4861609101295471, "epoch": 1.6205533596837944, "frac_reward_zero_std": 0.0, "grad_norm": 1.9517678022384644, "learning_rate": 2e-05, "loss": -7.0953369140625e-05, "num_tokens": 24537674.0, "reward": 1.0241973876953125, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 30.825, "reward/completion_length_min": 26.2, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.009375, "reward/duplicate_ranked_plain_query_rate": 0.009375, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.990625, "reward/mean_ndcg_at_10": 0.8953125, "reward/mean_r_ndcg": 0.7546875, "reward/mean_r_ndcg_cutoff": 6.63125, "reward/mean_recall": 0.221875, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.221875, "reward/mean_retrieval_combined_bounded": 0.0333404541015625, "reward/mean_retrieval_combined_raw": 0.11421691179275513, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 406.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3056685388088226, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.306640625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.0260774374008177, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.06635734997689724, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 13.488980484008788, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4472635924816132, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 410, "step_time": 13.811166894575582 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.2875, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.6, "completions/min_terminated_length": 0.0, "entropy": 0.5214999713003635, "epoch": 1.6403162055335967, "frac_reward_zero_std": 0.0, "grad_norm": 1.992564082145691, "learning_rate": 2e-05, "loss": -0.00150146484375, "num_tokens": 24833850.0, "reward": 1.035345458984375, "reward/completion_length_max": 34.6, "reward/completion_length_mean": 30.3, "reward/completion_length_min": 26.6, "reward/completion_length_p90": 32.4, "reward/completion_length_p95": 32.6, "reward/duplicate_query_rate": 0.0234375, "reward/duplicate_ranked_plain_query_rate": 0.0234375, "reward/invalid_output_rate": 0.0234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.034375, "reward/mean_colbert": 0.95234375, "reward/mean_mrr": 0.9765625, "reward/mean_ndcg_at_10": 0.865625, "reward/mean_r_ndcg": 0.740625, "reward/mean_r_ndcg_cutoff": 6.5875, "reward/mean_recall": 0.2400390625, "reward/mean_recall_at_10": 0.047705078125, "reward/mean_recall_at_100": 0.2400390625, "reward/mean_retrieval_combined_bounded": 0.0587890625, "reward/mean_retrieval_combined_raw": 0.163623046875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 412.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.31395490765571593, "rewards/retrieval_reward/mean": 1.034375, "rewards/retrieval_reward/std": 0.314453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3398143529891966, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.06854726225137711, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.14001235961914, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4619722306728363, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 415, "step_time": 13.647767053800635 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.1390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.4, "completions/min_terminated_length": 0.0, "entropy": 0.4514093965291977, "epoch": 1.6600790513833992, "frac_reward_zero_std": 0.0, "grad_norm": 2.3235039710998535, "learning_rate": 2e-05, "loss": -0.000634765625, "num_tokens": 25128716.0, "reward": 1.0321075439453125, "reward/completion_length_max": 33.8, "reward/completion_length_mean": 29.125, "reward/completion_length_min": 25.4, "reward/completion_length_p90": 31.2, "reward/completion_length_p95": 31.6, "reward/duplicate_query_rate": 0.01875, "reward/duplicate_ranked_plain_query_rate": 0.01875, "reward/invalid_output_rate": 0.01875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0328125, "reward/mean_colbert": 0.9578125, "reward/mean_mrr": 0.98125, "reward/mean_ndcg_at_10": 0.8546875, "reward/mean_r_ndcg": 0.73359375, "reward/mean_r_ndcg_cutoff": 6.475, "reward/mean_recall": 0.23984375, "reward/mean_recall_at_10": 0.04755859375, "reward/mean_recall_at_100": 0.23984375, "reward/mean_retrieval_combined_bounded": 0.050830078125, "reward/mean_retrieval_combined_raw": 0.13681676536798476, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 417.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2920800566673279, "rewards/retrieval_reward/mean": 1.0328125, "rewards/retrieval_reward/std": 0.2921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.9799049377441407, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.04791702069342137, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.034637069702148, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5247892260551452, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 420, "step_time": 13.604783405398484 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.5140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 23.2, "completions/min_terminated_length": 0.0, "entropy": 0.48933218494057656, "epoch": 1.6798418972332017, "frac_reward_zero_std": 0.0, "grad_norm": 2.157444715499878, "learning_rate": 2e-05, "loss": -0.00595703125, "num_tokens": 25422654.0, "reward": 0.9925750732421875, "reward/completion_length_max": 33.8, "reward/completion_length_mean": 28.525, "reward/completion_length_min": 23.2, "reward/completion_length_p90": 31.2, "reward/completion_length_p95": 31.8, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.03359375, "reward/length_out_of_range_rate": 0.02578125, "reward/low_length_query_rate": 0.02578125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99375, "reward/mean_colbert": 0.940625, "reward/mean_mrr": 0.96640625, "reward/mean_ndcg_at_10": 0.8765625, "reward/mean_r_ndcg": 0.74375, "reward/mean_r_ndcg_cutoff": 6.60625, "reward/mean_recall": 0.2201171875, "reward/mean_recall_at_10": 0.046728515625, "reward/mean_recall_at_100": 0.2201171875, "reward/mean_retrieval_combined_bounded": 0.04052734375, "reward/mean_retrieval_combined_raw": 0.07607468962669373, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.96640625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96640625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 422.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.02578125, "reward_std": 0.3774858295917511, "rewards/retrieval_reward/mean": 0.99375, "rewards/retrieval_reward/std": 0.37890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.8055246233940125, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.03020680509507656, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 16.806558227539064, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5528954267501831, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 425, "step_time": 13.54631095113 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.7703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.0, "completions/min_terminated_length": 0.0, "entropy": 0.48464804142713547, "epoch": 1.699604743083004, "frac_reward_zero_std": 0.0, "grad_norm": 2.00597882270813, "learning_rate": 2e-05, "loss": -0.001080322265625, "num_tokens": 25716792.0, "reward": 1.03336181640625, "reward/completion_length_max": 32.8, "reward/completion_length_mean": 28.75, "reward/completion_length_min": 25.0, "reward/completion_length_p90": 30.4, "reward/completion_length_p95": 31.4, "reward/duplicate_query_rate": 0.0109375, "reward/duplicate_ranked_plain_query_rate": 0.0109375, "reward/invalid_output_rate": 0.0171875, "reward/length_out_of_range_rate": 0.00625, "reward/low_length_query_rate": 0.00625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.034375, "reward/mean_colbert": 0.95859375, "reward/mean_mrr": 0.9828125, "reward/mean_ndcg_at_10": 0.86796875, "reward/mean_r_ndcg": 0.71640625, "reward/mean_r_ndcg_cutoff": 5.61875, "reward/mean_recall": 0.22109375, "reward/mean_recall_at_10": 0.04755859375, "reward/mean_recall_at_100": 0.22109375, "reward/mean_retrieval_combined_bounded": 0.0535400390625, "reward/mean_retrieval_combined_raw": 0.07617188093718141, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 426.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00625, "reward_std": 0.30260186195373534, "rewards/retrieval_reward/mean": 1.034375, "rewards/retrieval_reward/std": 0.303125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.049191403388977, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.04158458150923252, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.89242057800293, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5178352355957031, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 430, "step_time": 13.499693996459246 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.5921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.4, "completions/min_terminated_length": 0.0, "entropy": 0.45432404428720474, "epoch": 1.7193675889328062, "frac_reward_zero_std": 0.0, "grad_norm": 2.3616316318511963, "learning_rate": 2e-05, "loss": -0.000860595703125, "num_tokens": 26010926.0, "reward": 1.019696044921875, "reward/completion_length_max": 32.2, "reward/completion_length_mean": 28.575, "reward/completion_length_min": 25.4, "reward/completion_length_p90": 30.2, "reward/completion_length_p95": 31.0, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.86328125, "reward/mean_r_ndcg": 0.66796875, "reward/mean_r_ndcg_cutoff": 4.9625, "reward/mean_recall": 0.243359375, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.243359375, "reward/mean_retrieval_combined_bounded": 0.02254638671875, "reward/mean_retrieval_combined_raw": 0.02308236062526703, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 432.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27236926555633545, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.2720703125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.9963369607925414, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.03658425286412239, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 14.182538032531738, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5519958734512329, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 435, "step_time": 13.53439807976829 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.1984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.8, "completions/min_terminated_length": 0.0, "entropy": 0.47886601835489273, "epoch": 1.7391304347826086, "frac_reward_zero_std": 0.0, "grad_norm": 3.0130598545074463, "learning_rate": 2e-05, "loss": -0.00130615234375, "num_tokens": 26304492.0, "reward": 1.03223876953125, "reward/completion_length_max": 32.8, "reward/completion_length_mean": 28.2, "reward/completion_length_min": 24.8, "reward/completion_length_p90": 30.2, "reward/completion_length_p95": 30.2, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.0140625, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0328125, "reward/mean_colbert": 0.9625, "reward/mean_mrr": 0.98515625, "reward/mean_ndcg_at_10": 0.85625, "reward/mean_r_ndcg": 0.69375, "reward/mean_r_ndcg_cutoff": 5.91875, "reward/mean_recall": 0.2412109375, "reward/mean_recall_at_10": 0.048779296875, "reward/mean_recall_at_100": 0.2412109375, "reward/mean_retrieval_combined_bounded": 0.0482421875, "reward/mean_retrieval_combined_raw": 0.0667724609375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9859375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9859375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 437.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00390625, "reward_std": 0.3004114627838135, "rewards/retrieval_reward/mean": 1.0328125, "rewards/retrieval_reward/std": 0.30078125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.5561700344085694, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.027662184461951254, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 13.338196945190429, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5871975541114807, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 440, "step_time": 13.529342760005965 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.2109375, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.4, "completions/min_terminated_length": 0.0, "entropy": 0.4451370365917683, "epoch": 1.7588932806324111, "frac_reward_zero_std": 0.0, "grad_norm": 3.492856979370117, "learning_rate": 2e-05, "loss": -0.0048583984375, "num_tokens": 26598522.0, "reward": 1.019805908203125, "reward/completion_length_max": 34.4, "reward/completion_length_mean": 28.225, "reward/completion_length_min": 24.4, "reward/completion_length_p90": 31.0, "reward/completion_length_p95": 32.0, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.01640625, "reward/length_out_of_range_rate": 0.01171875, "reward/low_length_query_rate": 0.01171875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.02109375, "reward/mean_colbert": 0.9609375, "reward/mean_mrr": 0.98359375, "reward/mean_ndcg_at_10": 0.87265625, "reward/mean_r_ndcg": 0.778125, "reward/mean_r_ndcg_cutoff": 6.64375, "reward/mean_recall": 0.2380859375, "reward/mean_recall_at_10": 0.047705078125, "reward/mean_recall_at_100": 0.2380859375, "reward/mean_retrieval_combined_bounded": 0.042266845703125, "reward/mean_retrieval_combined_raw": 0.06582013741135598, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98359375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98359375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 442.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.01171875, "reward_std": 0.31892828941345214, "rewards/retrieval_reward/mean": 1.02109375, "rewards/retrieval_reward/std": 0.319140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.4978780388832091, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.017215026635676624, "sampling/raw_vllm_importance_sampling_ratio/min": 1.3112019446556382e-18, "sampling/raw_vllm_sampling_logp_difference/max": 14.884993171691894, "sampling/raw_vllm_sampling_logp_difference/mean": 0.7151676535606384, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 445, "step_time": 13.657954531244467 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.5046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.0, "completions/min_terminated_length": 0.0, "entropy": 0.4777214080095291, "epoch": 1.7786561264822134, "frac_reward_zero_std": 0.0, "grad_norm": 3.225482225418091, "learning_rate": 2e-05, "loss": -0.003466796875, "num_tokens": 26892448.0, "reward": 1.0166168212890625, "reward/completion_length_max": 34.2, "reward/completion_length_mean": 28.5, "reward/completion_length_min": 24.0, "reward/completion_length_p90": 31.0, "reward/completion_length_p95": 31.8, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.02109375, "reward/length_out_of_range_rate": 0.01328125, "reward/low_length_query_rate": 0.01328125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01796875, "reward/mean_colbert": 0.95625, "reward/mean_mrr": 0.97890625, "reward/mean_ndcg_at_10": 0.8734375, "reward/mean_r_ndcg": 0.7859375, "reward/mean_r_ndcg_cutoff": 6.7, "reward/mean_recall": 0.2140625, "reward/mean_recall_at_10": 0.0466796875, "reward/mean_recall_at_100": 0.2140625, "reward/mean_retrieval_combined_bounded": 0.0442138671875, "reward/mean_retrieval_combined_raw": 0.078125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 446.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.01328125, "reward_std": 0.3171050786972046, "rewards/retrieval_reward/mean": 1.01796875, "rewards/retrieval_reward/std": 0.31796875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.415211910009384, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.021446252753958107, "sampling/raw_vllm_importance_sampling_ratio/min": 2.1278035162441068e-16, "sampling/raw_vllm_sampling_logp_difference/max": 13.87791805267334, "sampling/raw_vllm_sampling_logp_difference/mean": 0.6854200959205627, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 450, "step_time": 13.581107450649142 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.09765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.0, "completions/min_terminated_length": 0.0, "entropy": 0.4908854402601719, "epoch": 1.7826086956521738, "frac_reward_zero_std": 0.0, "num_tokens": 26951769.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6511604613366067, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.701160427846325, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8769771528998243, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9956063268892794, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5529643244206083, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6075557116237167, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9733333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3207093948792668, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.359978827139009, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5277108433734939, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8409638554216867, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6120432254672278, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.622543290147757, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7535222699835376, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.8966255868544601, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5523571457318754, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.34135052634464674, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.167452346731272, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3013096966536911, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 0.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3452.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.3705759608740967, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4066900823694703, "retrieval_eval/nq/single_turn/recall_at_10": 0.6018250289687138, "retrieval_eval/nq/single_turn/recall_at_100": 0.8549633062958671, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6643402558970605, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6981747027051832, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8365555555555556, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9208888888888889, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.026153564453125, "reward/completion_length_max": 34.0, "reward/completion_length_mean": 30.125, "reward/completion_length_min": 26.0, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.8984375, "reward/mean_r_ndcg": 0.72265625, "reward/mean_r_ndcg_cutoff": 6.34375, "reward/mean_recall": 0.1611328125, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.1611328125, "reward/mean_retrieval_combined_bounded": 0.02978515625, "reward/mean_retrieval_combined_raw": 0.03173828125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 450.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.26954713463783264, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.26953125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 0.3551177382469177, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.0054103173315525055, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 14.481796264648438, "sampling/raw_vllm_sampling_logp_difference/mean": 0.7052998542785645, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 451, "step_time": 13.70749852102017 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.09765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.75, "completions/min_terminated_length": 0.0, "entropy": 0.5268598776310682, "epoch": 1.7984189723320159, "frac_reward_zero_std": 0.0, "grad_norm": 3.0223708152770996, "learning_rate": 2e-05, "loss": -0.00047607421875, "num_tokens": 27188733.0, "reward": 1.0097694396972656, "reward/completion_length_max": 35.0, "reward/completion_length_mean": 30.125, "reward/completion_length_min": 25.75, "reward/completion_length_p90": 32.5, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.0302734375, "reward/duplicate_ranked_plain_query_rate": 0.0302734375, "reward/invalid_output_rate": 0.0302734375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0087890625, "reward/mean_colbert": 0.9462890625, "reward/mean_mrr": 0.9697265625, "reward/mean_ndcg_at_10": 0.8525390625, "reward/mean_r_ndcg": 0.7744140625, "reward/mean_r_ndcg_cutoff": 5.8125, "reward/mean_recall": 0.25146484375, "reward/mean_recall_at_10": 0.04705810546875, "reward/mean_recall_at_100": 0.25146484375, "reward/mean_retrieval_combined_bounded": 0.0398406982421875, "reward/mean_retrieval_combined_raw": 0.10165715217590332, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9697265625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9697265625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 452.5, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3102407902479172, "rewards/retrieval_reward/mean": 1.0087890625, "rewards/retrieval_reward/std": 0.310546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 0.8977622911334038, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.014257345697842538, "sampling/raw_vllm_importance_sampling_ratio/min": 1.344011512078531e-23, "sampling/raw_vllm_sampling_logp_difference/max": 15.158933639526367, "sampling/raw_vllm_sampling_logp_difference/mean": 0.6437400281429291, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 455, "step_time": 13.727843130982365 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 56.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.88359375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.5486777395009994, "epoch": 1.8181818181818183, "frac_reward_zero_std": 0.0, "grad_norm": 2.3503599166870117, "learning_rate": 2e-05, "loss": 0.01708984375, "num_tokens": 27488360.0, "reward": 1.0041290283203126, "reward/completion_length_max": 56.0, "reward/completion_length_mean": 32.875, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 35.4, "reward/completion_length_p95": 36.2, "reward/duplicate_query_rate": 0.0359375, "reward/duplicate_ranked_plain_query_rate": 0.0359375, "reward/invalid_output_rate": 0.03828125, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00390625, "reward/mean_colbert": 0.93984375, "reward/mean_mrr": 0.96171875, "reward/mean_ndcg_at_10": 0.84921875, "reward/mean_r_ndcg": 0.70859375, "reward/mean_r_ndcg_cutoff": 5.8625, "reward/mean_recall": 0.223046875, "reward/mean_recall_at_10": 0.047021484375, "reward/mean_recall_at_100": 0.223046875, "reward/mean_retrieval_combined_bounded": 0.045846176147460935, "reward/mean_retrieval_combined_raw": 0.0801513671875, "reward/output_tag_rate": 0.9984375, "reward/query_valid_rate": 0.96171875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96171875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 457.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.32893542051315305, "rewards/retrieval_reward/mean": 1.00390625, "rewards/retrieval_reward/std": 0.329296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.2203951239585877, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.011769322399049997, "sampling/raw_vllm_importance_sampling_ratio/min": 2.6676835187481125e-20, "sampling/raw_vllm_sampling_logp_difference/max": 13.357924842834473, "sampling/raw_vllm_sampling_logp_difference/mean": 0.6458249807357788, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 460, "step_time": 14.69556798079284 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 59.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 35.09765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.0, "completions/min_terminated_length": 0.0, "entropy": 0.6061080932617188, "epoch": 1.8379446640316206, "frac_reward_zero_std": 0.0, "grad_norm": 2.1154110431671143, "learning_rate": 2e-05, "loss": 0.01494140625, "num_tokens": 27791141.0, "reward": 1.03953857421875, "reward/completion_length_max": 59.6, "reward/completion_length_mean": 35.05, "reward/completion_length_min": 30.0, "reward/completion_length_p90": 38.4, "reward/completion_length_p95": 39.2, "reward/duplicate_query_rate": 0.02265625, "reward/duplicate_ranked_plain_query_rate": 0.02265625, "reward/invalid_output_rate": 0.0390625, "reward/length_out_of_range_rate": 0.015625, "reward/low_length_query_rate": 0.015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.040625, "reward/mean_colbert": 0.93828125, "reward/mean_mrr": 0.9609375, "reward/mean_ndcg_at_10": 0.8828125, "reward/mean_r_ndcg": 0.7671875, "reward/mean_r_ndcg_cutoff": 6.81875, "reward/mean_recall": 0.2103515625, "reward/mean_recall_at_10": 0.04677734375, "reward/mean_recall_at_100": 0.2103515625, "reward/mean_retrieval_combined_bounded": 0.087890625, "reward/mean_retrieval_combined_raw": 0.18017578125, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.9609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 462.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.35247358679771423, "rewards/retrieval_reward/mean": 1.040625, "rewards/retrieval_reward/std": 0.353515625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.7794322848320008, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.011306994687765836, "sampling/raw_vllm_importance_sampling_ratio/min": 1.3169820845743015e-22, "sampling/raw_vllm_sampling_logp_difference/max": 11.733940124511719, "sampling/raw_vllm_sampling_logp_difference/mean": 0.669526469707489, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 465, "step_time": 14.896740095550195 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 60.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.90703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.6, "completions/min_terminated_length": 0.0, "entropy": 0.5700881019234657, "epoch": 1.8577075098814229, "frac_reward_zero_std": 0.0, "grad_norm": 2.2216062545776367, "learning_rate": 2e-05, "loss": 0.0130859375, "num_tokens": 28093486.0, "reward": 0.996051025390625, "reward/completion_length_max": 60.4, "reward/completion_length_mean": 34.9, "reward/completion_length_min": 29.6, "reward/completion_length_p90": 38.2, "reward/completion_length_p95": 39.2, "reward/duplicate_query_rate": 0.009375, "reward/duplicate_ranked_plain_query_rate": 0.009375, "reward/invalid_output_rate": 0.0359375, "reward/length_out_of_range_rate": 0.025, "reward/low_length_query_rate": 0.025, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99453125, "reward/mean_colbert": 0.93984375, "reward/mean_mrr": 0.9640625, "reward/mean_ndcg_at_10": 0.8921875, "reward/mean_r_ndcg": 0.7765625, "reward/mean_r_ndcg_cutoff": 6.21875, "reward/mean_recall": 0.2408203125, "reward/mean_recall_at_10": 0.04736328125, "reward/mean_recall_at_100": 0.2408203125, "reward/mean_retrieval_combined_bounded": 0.047705078125, "reward/mean_retrieval_combined_raw": 0.114794921875, "reward/output_tag_rate": 0.9984375, "reward/query_valid_rate": 0.9640625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9640625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 466.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.025, "reward/too_short_query_rate": 0.0, "reward_std": 0.38973424434661863, "rewards/retrieval_reward/mean": 0.99453125, "rewards/retrieval_reward/std": 0.38984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.3581327676773072, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.012648508884012699, "sampling/raw_vllm_importance_sampling_ratio/min": 9.669519119164594e-18, "sampling/raw_vllm_sampling_logp_difference/max": 13.279877281188964, "sampling/raw_vllm_sampling_logp_difference/mean": 0.6000359177589416, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 470, "step_time": 14.969685159204527 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 57.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.22421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.2, "completions/min_terminated_length": 0.0, "entropy": 0.5589335545897484, "epoch": 1.8774703557312253, "frac_reward_zero_std": 0.0, "grad_norm": 1.7110692262649536, "learning_rate": 2e-05, "loss": 0.0130859375, "num_tokens": 28393389.0, "reward": 1.0325042724609375, "reward/completion_length_max": 57.8, "reward/completion_length_mean": 33.25, "reward/completion_length_min": 27.2, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 36.8, "reward/duplicate_query_rate": 0.0109375, "reward/duplicate_ranked_plain_query_rate": 0.0109375, "reward/invalid_output_rate": 0.0140625, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.034375, "reward/mean_colbert": 0.9640625, "reward/mean_mrr": 0.9859375, "reward/mean_ndcg_at_10": 0.8984375, "reward/mean_r_ndcg": 0.8171875, "reward/mean_r_ndcg_cutoff": 6.7375, "reward/mean_recall": 0.216796875, "reward/mean_recall_at_10": 0.0484375, "reward/mean_recall_at_100": 0.216796875, "reward/mean_retrieval_combined_bounded": 0.04921875, "reward/mean_retrieval_combined_raw": 0.085205078125, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.9859375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9859375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 472.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.29071147441864015, "rewards/retrieval_reward/mean": 1.034375, "rewards/retrieval_reward/std": 0.291015625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.381622815132141, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.029499047435820103, "sampling/raw_vllm_importance_sampling_ratio/min": 1.1846111863558e-15, "sampling/raw_vllm_sampling_logp_difference/max": 13.554391479492187, "sampling/raw_vllm_sampling_logp_difference/mean": 0.54606032371521, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 475, "step_time": 14.792859670019244 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.73828125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.8, "completions/min_terminated_length": 0.0, "entropy": 0.5544483706355094, "epoch": 1.8972332015810278, "frac_reward_zero_std": 0.0, "grad_norm": 2.1595561504364014, "learning_rate": 2e-05, "loss": 0.0011962890625, "num_tokens": 28692990.0, "reward": 1.0249603271484375, "reward/completion_length_max": 40.6, "reward/completion_length_mean": 32.75, "reward/completion_length_min": 27.8, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.9640625, "reward/mean_mrr": 0.98984375, "reward/mean_ndcg_at_10": 0.8578125, "reward/mean_r_ndcg": 0.71171875, "reward/mean_r_ndcg_cutoff": 5.25625, "reward/mean_recall": 0.2166015625, "reward/mean_recall_at_10": 0.047265625, "reward/mean_recall_at_100": 0.2166015625, "reward/mean_retrieval_combined_bounded": 0.03516845703125, "reward/mean_retrieval_combined_raw": 0.0629150390625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 477.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.27551308274269104, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.275390625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.620930653810501, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.03194218017160892, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.371166038513184, "sampling/raw_vllm_sampling_logp_difference/mean": 0.5357925593852997, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 480, "step_time": 14.012996836588718 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.93359375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.5373702712357045, "epoch": 1.91699604743083, "frac_reward_zero_std": 0.0, "grad_norm": 1.9640607833862305, "learning_rate": 2e-05, "loss": -0.00074462890625, "num_tokens": 28991625.0, "reward": 1.0034881591796876, "reward/completion_length_max": 39.4, "reward/completion_length_mean": 31.9, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.0203125, "reward/length_out_of_range_rate": 0.0125, "reward/low_length_query_rate": 0.0125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0046875, "reward/mean_colbert": 0.9546875, "reward/mean_mrr": 0.9796875, "reward/mean_ndcg_at_10": 0.8625, "reward/mean_r_ndcg": 0.71171875, "reward/mean_r_ndcg_cutoff": 5.625, "reward/mean_recall": 0.221875, "reward/mean_recall_at_10": 0.048486328125, "reward/mean_recall_at_100": 0.221875, "reward/mean_retrieval_combined_bounded": 0.03173828125, "reward/mean_retrieval_combined_raw": 0.07251146556809544, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.9796875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9796875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 482.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0125, "reward/too_short_query_rate": 0.0, "reward_std": 0.33292731642723083, "rewards/retrieval_reward/mean": 1.003125, "rewards/retrieval_reward/std": 0.333203125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.261406898498535, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.06279257163405419, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.699299621582032, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4345217525959015, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 485, "step_time": 13.92449243098963 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.53359375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.4854669362306595, "epoch": 1.9367588932806323, "frac_reward_zero_std": 0.0, "grad_norm": 2.6642093658447266, "learning_rate": 2e-05, "loss": -0.00189208984375, "num_tokens": 29288276.0, "reward": 1.012384033203125, "reward/completion_length_max": 36.4, "reward/completion_length_mean": 30.55, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 33.4, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01484375, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.90703125, "reward/mean_r_ndcg": 0.7859375, "reward/mean_r_ndcg_cutoff": 6.73125, "reward/mean_recall": 0.25703125, "reward/mean_recall_at_10": 0.04853515625, "reward/mean_recall_at_100": 0.25703125, "reward/mean_retrieval_combined_bounded": 0.01549072265625, "reward/mean_retrieval_combined_raw": 0.016113820672035217, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 486.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2818064332008362, "rewards/retrieval_reward/mean": 1.01484375, "rewards/retrieval_reward/std": 0.281640625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4681758642196656, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.0955021008849144, "sampling/raw_vllm_importance_sampling_ratio/min": 2.2389947155424667e-14, "sampling/raw_vllm_sampling_logp_difference/max": 10.374510192871094, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3789074242115021, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 490, "step_time": 13.766250502644107 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.525, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.0, "completions/min_terminated_length": 0.0, "entropy": 0.4524750828742981, "epoch": 1.9565217391304348, "frac_reward_zero_std": 0.0, "grad_norm": 1.616303563117981, "learning_rate": 2e-05, "loss": -0.00101318359375, "num_tokens": 29584628.0, "reward": 1.0027984619140624, "reward/completion_length_max": 35.0, "reward/completion_length_mean": 30.525, "reward/completion_length_min": 25.0, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 32.8, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00234375, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.99140625, "reward/mean_ndcg_at_10": 0.90234375, "reward/mean_r_ndcg": 0.775, "reward/mean_r_ndcg_cutoff": 6.39375, "reward/mean_recall": 0.2171875, "reward/mean_recall_at_10": 0.0482421875, "reward/mean_recall_at_100": 0.2171875, "reward/mean_retrieval_combined_bounded": 0.0051513671875, "reward/mean_retrieval_combined_raw": 0.006410005316138267, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 492.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.29597532749176025, "rewards/retrieval_reward/mean": 1.00234375, "rewards/retrieval_reward/std": 0.296484375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.469842958450317, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.0844101920723915, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.52134017944336, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3653779625892639, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 495, "step_time": 13.733681112935301 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 43.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.09765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.48316261544823647, "epoch": 1.9762845849802373, "frac_reward_zero_std": 0.0, "grad_norm": 1.728529691696167, "learning_rate": 2e-05, "loss": 0.0042724609375, "num_tokens": 29881841.0, "reward": 1.0241912841796874, "reward/completion_length_max": 43.4, "reward/completion_length_mean": 31.075, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.8953125, "reward/mean_r_ndcg": 0.7953125, "reward/mean_r_ndcg_cutoff": 6.55, "reward/mean_recall": 0.2294921875, "reward/mean_recall_at_10": 0.0486328125, "reward/mean_recall_at_100": 0.2294921875, "reward/mean_retrieval_combined_bounded": 0.0318603515625, "reward/mean_retrieval_combined_raw": 0.03344801962375641, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 497.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2846437573432922, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.284765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.9844703912734984, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.059231369942426684, "sampling/raw_vllm_importance_sampling_ratio/min": 1.0097668361186417e-12, "sampling/raw_vllm_sampling_logp_difference/max": 9.642176818847656, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4092893719673157, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 500, "step_time": 14.111372668214608 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.85546875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.4324837028980255, "epoch": 1.9802371541501977, "frac_reward_zero_std": 0.0, "num_tokens": 29941292.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6178210806094037, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.6742020420695034, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8717047451669596, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9964850615114236, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6004471034071244, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.655489900356949, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8466666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3315223485713352, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3753694445677392, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5566265060240964, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8626506024096385, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6041355663718626, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6126974443286604, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7342357173343089, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9067823608316566, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5831825541131028, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3584882345509702, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17442117550937794, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3161678540056533, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 1.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3451.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.40329294845917874, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4404731467948958, "retrieval_eval/nq/single_turn/recall_at_10": 0.6402974983096686, "retrieval_eval/nq/single_turn/recall_at_100": 0.877740751473003, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.659214876586842, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6925224687703636, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8283333333333334, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9358888888888889, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0363006591796875, "reward/completion_length_max": 35.0, "reward/completion_length_mean": 30.875, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.01171875, "reward/duplicate_ranked_plain_query_rate": 0.01171875, "reward/invalid_output_rate": 0.01171875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0390625, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.984375, "reward/mean_ndcg_at_10": 0.875, "reward/mean_r_ndcg": 0.60546875, "reward/mean_r_ndcg_cutoff": 5.0625, "reward/mean_recall": 0.28515625, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.28515625, "reward/mean_retrieval_combined_bounded": 0.0478515625, "reward/mean_retrieval_combined_raw": 0.07421875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 500.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2577993869781494, "rewards/retrieval_reward/mean": 1.0390625, "rewards/retrieval_reward/std": 0.2578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.122814178466797, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.07863526046276093, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.52499008178711, "sampling/raw_vllm_sampling_logp_difference/mean": 0.36716872453689575, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 501, "step_time": 13.710477588989306 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 61.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.9345703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.5, "completions/min_terminated_length": 0.0, "entropy": 0.4565559634938836, "epoch": 1.9960474308300395, "frac_reward_zero_std": 0.0, "grad_norm": 1.7587902545928955, "learning_rate": 2e-05, "loss": 0.01171875, "num_tokens": 30180233.0, "reward": 1.0209770202636719, "reward/completion_length_max": 61.25, "reward/completion_length_mean": 32.0, "reward/completion_length_min": 27.5, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.5, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.0107421875, "reward/length_out_of_range_rate": 0.001953125, "reward/low_length_query_rate": 0.001953125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021484375, "reward/mean_colbert": 0.9658203125, "reward/mean_mrr": 0.9892578125, "reward/mean_ndcg_at_10": 0.9033203125, "reward/mean_r_ndcg": 0.7724609375, "reward/mean_r_ndcg_cutoff": 6.0859375, "reward/mean_recall": 0.23388671875, "reward/mean_recall_at_10": 0.0489501953125, "reward/mean_recall_at_100": 0.23388671875, "reward/mean_retrieval_combined_bounded": 0.03482818603515625, "reward/mean_retrieval_combined_raw": 0.05016550421714783, "reward/output_tag_rate": 0.9990234375, "reward/query_valid_rate": 0.9892578125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9892578125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 502.5, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.001953125, "reward/too_short_query_rate": 0.0, "reward_std": 0.30356816947460175, "rewards/retrieval_reward/mean": 1.021484375, "rewards/retrieval_reward/std": 0.3037109375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.522892028093338, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.09280001744627953, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.644492864608765, "sampling/raw_vllm_sampling_logp_difference/mean": 0.36969956010580063, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 505, "step_time": 14.939245536763337 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.33515625, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.49822128787636755, "epoch": 2.015810276679842, "frac_reward_zero_std": 0.0, "grad_norm": 1.5699563026428223, "learning_rate": 2e-05, "loss": -0.001251220703125, "num_tokens": 30477846.0, "reward": 1.0139892578125, "reward/completion_length_max": 36.8, "reward/completion_length_mean": 31.3, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 33.8, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.99765625, "reward/mean_ndcg_at_10": 0.89765625, "reward/mean_r_ndcg": 0.70078125, "reward/mean_r_ndcg_cutoff": 6.28125, "reward/mean_recall": 0.2423828125, "reward/mean_recall_at_10": 0.048388671875, "reward/mean_recall_at_100": 0.2423828125, "reward/mean_retrieval_combined_bounded": 0.016460418701171875, "reward/mean_retrieval_combined_raw": 0.018778014183044433, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 506.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28199478387832644, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.28203125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.58535475730896, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.08535245284438134, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.793828773498536, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3562849223613739, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 510, "step_time": 13.849758668185677 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.76328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.5376276500523091, "epoch": 2.035573122529644, "frac_reward_zero_std": 0.0, "grad_norm": 1.5375944375991821, "learning_rate": 2e-05, "loss": 0.000848388671875, "num_tokens": 30775687.0, "reward": 1.0262725830078125, "reward/completion_length_max": 36.8, "reward/completion_length_mean": 31.725, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.9125, "reward/mean_r_ndcg": 0.778125, "reward/mean_r_ndcg_cutoff": 6.275, "reward/mean_recall": 0.23828125, "reward/mean_recall_at_10": 0.048388671875, "reward/mean_recall_at_100": 0.23828125, "reward/mean_retrieval_combined_bounded": 0.030078125, "reward/mean_retrieval_combined_raw": 0.030671211332082747, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 511.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.269369101524353, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.269921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6805997848510743, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.14325588047504426, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.571949577331543, "sampling/raw_vllm_sampling_logp_difference/mean": 0.31887912154197695, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 515, "step_time": 13.80650453295093 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.0515625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.4915900588035583, "epoch": 2.0553359683794468, "frac_reward_zero_std": 0.0, "grad_norm": 1.7367616891860962, "learning_rate": 2e-05, "loss": 0.00257568359375, "num_tokens": 31074345.0, "reward": 1.0193756103515625, "reward/completion_length_max": 38.8, "reward/completion_length_mean": 32.05, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 33.8, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.0171875, "reward/duplicate_ranked_plain_query_rate": 0.0171875, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.95546875, "reward/mean_mrr": 0.98046875, "reward/mean_ndcg_at_10": 0.87109375, "reward/mean_r_ndcg": 0.71484375, "reward/mean_r_ndcg_cutoff": 5.8375, "reward/mean_recall": 0.2337890625, "reward/mean_recall_at_10": 0.047900390625, "reward/mean_recall_at_100": 0.2337890625, "reward/mean_retrieval_combined_bounded": 0.040087890625, "reward/mean_retrieval_combined_raw": 0.06538103483617305, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 517.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.3130422294139862, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.313671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5273618936538695, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1080674558877945, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 13.470388603210449, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3756819427013397, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 520, "step_time": 13.893956706253812 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.2328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.2, "completions/min_terminated_length": 0.0, "entropy": 0.5217269696295261, "epoch": 2.075098814229249, "frac_reward_zero_std": 0.0, "grad_norm": 1.7329813241958618, "learning_rate": 2e-05, "loss": 0.0029052734375, "num_tokens": 31374323.0, "reward": 1.0254364013671875, "reward/completion_length_max": 41.6, "reward/completion_length_mean": 33.2, "reward/completion_length_min": 29.2, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.009375, "reward/duplicate_ranked_plain_query_rate": 0.009375, "reward/invalid_output_rate": 0.015625, "reward/length_out_of_range_rate": 0.00625, "reward/low_length_query_rate": 0.00625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.96015625, "reward/mean_mrr": 0.9828125, "reward/mean_ndcg_at_10": 0.8921875, "reward/mean_r_ndcg": 0.68359375, "reward/mean_r_ndcg_cutoff": 6.45625, "reward/mean_recall": 0.2224609375, "reward/mean_recall_at_10": 0.048046875, "reward/mean_recall_at_100": 0.2224609375, "reward/mean_retrieval_combined_bounded": 0.044036865234375, "reward/mean_retrieval_combined_raw": 0.08857421875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 521.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00625, "reward/too_short_query_rate": 0.0, "reward_std": 0.29462796449661255, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.295703125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8255728244781495, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.11547213643789292, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.410499954223633, "sampling/raw_vllm_sampling_logp_difference/mean": 0.32353546619415285, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 525, "step_time": 14.025626784202178 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.62734375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.5224706396460533, "epoch": 2.0948616600790513, "frac_reward_zero_std": 0.0, "grad_norm": 1.9015051126480103, "learning_rate": 2e-05, "loss": 0.001092529296875, "num_tokens": 31673750.0, "reward": 1.00850830078125, "reward/completion_length_max": 40.2, "reward/completion_length_mean": 32.65, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 36.2, "reward/duplicate_query_rate": 0.01484375, "reward/duplicate_ranked_plain_query_rate": 0.01484375, "reward/invalid_output_rate": 0.01796875, "reward/length_out_of_range_rate": 0.003125, "reward/low_length_query_rate": 0.003125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00859375, "reward/mean_colbert": 0.95703125, "reward/mean_mrr": 0.98203125, "reward/mean_ndcg_at_10": 0.865625, "reward/mean_r_ndcg": 0.74453125, "reward/mean_r_ndcg_cutoff": 5.79375, "reward/mean_recall": 0.226953125, "reward/mean_recall_at_10": 0.048193359375, "reward/mean_recall_at_100": 0.226953125, "reward/mean_retrieval_combined_bounded": 0.027996826171875, "reward/mean_retrieval_combined_raw": 0.06479699164628983, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98203125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98203125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 527.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.003125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3110500335693359, "rewards/retrieval_reward/mean": 1.00859375, "rewards/retrieval_reward/std": 0.311328125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.321122074127197, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.09599485099315644, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.921562194824219, "sampling/raw_vllm_sampling_logp_difference/mean": 0.35063697695732116, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 530, "step_time": 13.956009857403114 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.16875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.4891761511564255, "epoch": 2.1146245059288535, "frac_reward_zero_std": 0.0, "grad_norm": 1.588834524154663, "learning_rate": 2e-05, "loss": 0.001385498046875, "num_tokens": 31972110.0, "reward": 1.031103515625, "reward/completion_length_max": 39.4, "reward/completion_length_mean": 32.175, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.0046875, "reward/low_length_query_rate": 0.0046875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.028125, "reward/mean_colbert": 0.965625, "reward/mean_mrr": 0.990625, "reward/mean_ndcg_at_10": 0.875, "reward/mean_r_ndcg": 0.78671875, "reward/mean_r_ndcg_cutoff": 6.31875, "reward/mean_recall": 0.2548828125, "reward/mean_recall_at_10": 0.0486328125, "reward/mean_recall_at_100": 0.2548828125, "reward/mean_retrieval_combined_bounded": 0.042822265625, "reward/mean_retrieval_combined_raw": 0.051659274473786355, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 532.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0046875, "reward/too_short_query_rate": 0.0, "reward_std": 0.29105560183525087, "rewards/retrieval_reward/mean": 1.0296875, "rewards/retrieval_reward/std": 0.29140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.603234815597534, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1538059577345848, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.880440711975098, "sampling/raw_vllm_sampling_logp_difference/mean": 0.28332419991493224, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 535, "step_time": 13.885179100802635 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.51328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.8, "completions/min_terminated_length": 0.0, "entropy": 0.5067706666886806, "epoch": 2.1343873517786562, "frac_reward_zero_std": 0.0, "grad_norm": 1.6157203912734985, "learning_rate": 2e-05, "loss": -0.00223388671875, "num_tokens": 32270271.0, "reward": 1.0075103759765625, "reward/completion_length_max": 37.4, "reward/completion_length_mean": 31.5, "reward/completion_length_min": 27.8, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.0109375, "reward/duplicate_ranked_plain_query_rate": 0.0109375, "reward/invalid_output_rate": 0.0109375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0078125, "reward/mean_colbert": 0.965625, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.90546875, "reward/mean_r_ndcg": 0.76640625, "reward/mean_r_ndcg_cutoff": 6.44375, "reward/mean_recall": 0.2283203125, "reward/mean_recall_at_10": 0.047900390625, "reward/mean_recall_at_100": 0.2283203125, "reward/mean_retrieval_combined_bounded": 0.01856536865234375, "reward/mean_retrieval_combined_raw": 0.024453020095825194, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 536.8, "reward/search_failure_rate": 0.00078125, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2947881281375885, "rewards/retrieval_reward/mean": 1.0078125, "rewards/retrieval_reward/std": 0.294921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.659325933456421, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16951057314872742, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.256407451629638, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2777422577142715, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 540, "step_time": 13.780514171009418 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.71796875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.8, "completions/min_terminated_length": 0.0, "entropy": 0.5238783977925777, "epoch": 2.1541501976284585, "frac_reward_zero_std": 0.0, "grad_norm": 1.3645415306091309, "learning_rate": 2e-05, "loss": -0.00076904296875, "num_tokens": 32568278.0, "reward": 1.00928955078125, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 31.75, "reward/completion_length_min": 27.8, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.03828125, "reward/duplicate_ranked_plain_query_rate": 0.03828125, "reward/invalid_output_rate": 0.03828125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.009375, "reward/mean_colbert": 0.9359375, "reward/mean_mrr": 0.96171875, "reward/mean_ndcg_at_10": 0.8640625, "reward/mean_r_ndcg": 0.74921875, "reward/mean_r_ndcg_cutoff": 5.70625, "reward/mean_recall": 0.2177734375, "reward/mean_recall_at_10": 0.047216796875, "reward/mean_recall_at_100": 0.2177734375, "reward/mean_retrieval_combined_bounded": 0.04765625, "reward/mean_retrieval_combined_raw": 0.08137588500976563, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.96171875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96171875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 542.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3271628737449646, "rewards/retrieval_reward/mean": 1.009375, "rewards/retrieval_reward/std": 0.327734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7872326374053955, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.14923461228609086, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.959915161132812, "sampling/raw_vllm_sampling_logp_difference/mean": 0.30090678334236143, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 545, "step_time": 13.8532050107955 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.8375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.8, "completions/min_terminated_length": 0.0, "entropy": 0.5214081533253193, "epoch": 2.1739130434782608, "frac_reward_zero_std": 0.0, "grad_norm": 1.5344618558883667, "learning_rate": 2e-05, "loss": 0.00228271484375, "num_tokens": 32867718.0, "reward": 1.023333740234375, "reward/completion_length_max": 39.2, "reward/completion_length_mean": 32.825, "reward/completion_length_min": 28.8, "reward/completion_length_p90": 34.8, "reward/completion_length_p95": 35.4, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.00859375, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.965625, "reward/mean_mrr": 0.99140625, "reward/mean_ndcg_at_10": 0.89375, "reward/mean_r_ndcg": 0.7203125, "reward/mean_r_ndcg_cutoff": 5.74375, "reward/mean_recall": 0.2212890625, "reward/mean_recall_at_10": 0.04873046875, "reward/mean_recall_at_100": 0.2212890625, "reward/mean_retrieval_combined_bounded": 0.033837890625, "reward/mean_retrieval_combined_raw": 0.1167972020804882, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99140625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99140625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 546.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00390625, "reward/too_short_query_rate": 0.0, "reward_std": 0.29767284989356996, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.2984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8053430557250976, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.11116571128368377, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.851389122009277, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3196537375450134, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 550, "step_time": 13.916345103189816 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.8125, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.0, "completions/min_terminated_length": 0.0, "entropy": 0.5249018520116806, "epoch": 2.177865612648221, "frac_reward_zero_std": 0.0, "num_tokens": 32927862.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6076689978056565, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.663794996278636, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8629173989455184, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9973637961335676, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6044161840430237, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6608396539584046, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8533333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3369202066083638, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3741791545169944, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5397590361445783, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8626506024096385, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.5950852872807675, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6001711207384993, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7265547832449241, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9090124077800135, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5769460630197804, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3619277725281797, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.18075133237922333, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3216887441301572, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 1.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3451.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.4025518046077474, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4403836828628569, "retrieval_eval/nq/single_turn/recall_at_10": 0.6402492031295277, "retrieval_eval/nq/single_turn/recall_at_100": 0.883125664058727, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6676787090320262, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6997399535506199, "retrieval_eval/scifact/single_turn/recall_at_10": 0.834, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9426666666666667, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0156707763671875, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 33.75, "reward/completion_length_min": 31.0, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.9765625, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.921875, "reward/mean_r_ndcg": 0.7734375, "reward/mean_r_ndcg_cutoff": 6.875, "reward/mean_recall": 0.2041015625, "reward/mean_recall_at_10": 0.04931640625, "reward/mean_recall_at_100": 0.2041015625, "reward/mean_retrieval_combined_bounded": 0.015625, "reward/mean_retrieval_combined_raw": 2.8908252716064453e-06, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 552.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.29563775658607483, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.294921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.9590792655944824, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.11110767722129822, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.833253860473633, "sampling/raw_vllm_sampling_logp_difference/mean": 0.30866947770118713, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 551, "step_time": 13.861028263985645 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 43.5, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.54296875, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.75, "completions/min_terminated_length": 0.0, "entropy": 0.5615814421325922, "epoch": 2.1936758893280635, "frac_reward_zero_std": 0.0, "grad_norm": 1.4394816160202026, "learning_rate": 2e-05, "loss": 0.0053955078125, "num_tokens": 33169538.0, "reward": 0.9612808227539062, "reward/completion_length_max": 43.5, "reward/completion_length_mean": 34.5, "reward/completion_length_min": 29.75, "reward/completion_length_p90": 38.5, "reward/completion_length_p95": 40.0, "reward/duplicate_query_rate": 0.0107421875, "reward/duplicate_ranked_plain_query_rate": 0.0107421875, "reward/invalid_output_rate": 0.064453125, "reward/length_out_of_range_rate": 0.0458984375, "reward/low_length_query_rate": 0.0458984375, "reward/malformed_query_rate": 0.0068359375, "reward/malformed_ranked_plain_query_rate": 0.0068359375, "reward/mean": 0.9599609375, "reward/mean_colbert": 0.9111328125, "reward/mean_mrr": 0.935546875, "reward/mean_ndcg_at_10": 0.8193359375, "reward/mean_r_ndcg": 0.67578125, "reward/mean_r_ndcg_cutoff": 5.375, "reward/mean_recall": 0.212158203125, "reward/mean_recall_at_10": 0.0452880859375, "reward/mean_recall_at_100": 0.212158203125, "reward/mean_retrieval_combined_bounded": 0.057769775390625, "reward/mean_retrieval_combined_raw": 0.1378173828125, "reward/output_tag_rate": 0.9990234375, "reward/query_valid_rate": 0.935546875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.935546875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 552.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0458984375, "reward/too_short_query_rate": 0.0, "reward_std": 0.4498700425028801, "rewards/retrieval_reward/mean": 0.9599609375, "rewards/retrieval_reward/std": 0.451171875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3212271630764008, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.10690084472298622, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.501043796539307, "sampling/raw_vllm_sampling_logp_difference/mean": 0.29870715737342834, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 555, "step_time": 14.15134708005644 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.02421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.4, "completions/min_terminated_length": 0.0, "entropy": 0.5667620211839676, "epoch": 2.2134387351778657, "frac_reward_zero_std": 0.0, "grad_norm": 1.6278821229934692, "learning_rate": 2e-05, "loss": 0.0009765625, "num_tokens": 33470497.0, "reward": 0.9999755859375, "reward/completion_length_max": 40.0, "reward/completion_length_mean": 34.05, "reward/completion_length_min": 29.4, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 37.4, "reward/duplicate_query_rate": 0.02265625, "reward/duplicate_ranked_plain_query_rate": 0.02265625, "reward/invalid_output_rate": 0.02734375, "reward/length_out_of_range_rate": 0.0046875, "reward/low_length_query_rate": 0.0046875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99921875, "reward/mean_colbert": 0.94765625, "reward/mean_mrr": 0.97265625, "reward/mean_ndcg_at_10": 0.8515625, "reward/mean_r_ndcg": 0.73828125, "reward/mean_r_ndcg_cutoff": 6.59375, "reward/mean_recall": 0.2537109375, "reward/mean_recall_at_10": 0.04765625, "reward/mean_recall_at_100": 0.2537109375, "reward/mean_retrieval_combined_bounded": 0.029736328125, "reward/mean_retrieval_combined_raw": -0.00241546630859375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97265625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97265625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 557.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0046875, "reward/too_short_query_rate": 0.0, "reward_std": 0.3285679340362549, "rewards/retrieval_reward/mean": 0.99921875, "rewards/retrieval_reward/std": 0.329296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.1594780683517456, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.10015447735786438, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.057708358764648, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3084015786647797, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 560, "step_time": 13.992322540003807 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 44.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.3328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.0, "completions/min_terminated_length": 0.0, "entropy": 0.5749719843268395, "epoch": 2.233201581027668, "frac_reward_zero_std": 0.0, "grad_norm": 1.8580559492111206, "learning_rate": 2e-05, "loss": 0.0056884765625, "num_tokens": 33770379.0, "reward": 1.0189300537109376, "reward/completion_length_max": 44.6, "reward/completion_length_mean": 33.3, "reward/completion_length_min": 30.0, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 36.6, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.01328125, "reward/length_out_of_range_rate": 0.00546875, "reward/low_length_query_rate": 0.00546875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.9625, "reward/mean_mrr": 0.98671875, "reward/mean_ndcg_at_10": 0.85703125, "reward/mean_r_ndcg": 0.70703125, "reward/mean_r_ndcg_cutoff": 6.5625, "reward/mean_recall": 0.2275390625, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.2275390625, "reward/mean_retrieval_combined_bounded": 0.03497023582458496, "reward/mean_retrieval_combined_raw": 0.05830081906169653, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98671875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98671875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 561.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00546875, "reward/too_short_query_rate": 0.0, "reward_std": 0.2973390698432922, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.29765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.514517617225647, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1405956745147705, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.674882507324218, "sampling/raw_vllm_sampling_logp_difference/mean": 0.31269183158874514, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 565, "step_time": 14.192667050799354 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.8421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.6203115358948708, "epoch": 2.2529644268774702, "frac_reward_zero_std": 0.0, "grad_norm": 1.7430702447891235, "learning_rate": 2e-05, "loss": 0.00234375, "num_tokens": 34071329.0, "reward": 1.01448974609375, "reward/completion_length_max": 39.6, "reward/completion_length_mean": 33.85, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 37.2, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.003125, "reward/low_length_query_rate": 0.003125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.99453125, "reward/mean_ndcg_at_10": 0.88671875, "reward/mean_r_ndcg": 0.74921875, "reward/mean_r_ndcg_cutoff": 6.3875, "reward/mean_recall": 0.234765625, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.234765625, "reward/mean_retrieval_combined_bounded": 0.022216796875, "reward/mean_retrieval_combined_raw": 0.029711627028882505, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 567.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.003125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2875814467668533, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.287890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5887239456176756, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12464963793754577, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.87583656311035, "sampling/raw_vllm_sampling_logp_difference/mean": 0.31323943138122556, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 570, "step_time": 13.963273352070246 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.9046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.5567676283419132, "epoch": 2.2727272727272725, "frac_reward_zero_std": 0.0, "grad_norm": 1.4305529594421387, "learning_rate": 2e-05, "loss": -0.0019775390625, "num_tokens": 34370823.0, "reward": 1.00347900390625, "reward/completion_length_max": 39.4, "reward/completion_length_mean": 32.9, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.8, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00390625, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.896875, "reward/mean_r_ndcg": 0.7953125, "reward/mean_r_ndcg_cutoff": 6.16875, "reward/mean_recall": 0.2419921875, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.2419921875, "reward/mean_retrieval_combined_bounded": 0.0077056884765625, "reward/mean_retrieval_combined_raw": -0.011383247375488282, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 572.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.28356608748435974, "rewards/retrieval_reward/mean": 1.00390625, "rewards/retrieval_reward/std": 0.28359375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7729361057281494, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13580617010593415, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.713996887207031, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2862787365913391, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 575, "step_time": 13.90512731699273 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.99140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.5732641376554966, "epoch": 2.292490118577075, "frac_reward_zero_std": 0.0, "grad_norm": 1.373497486114502, "learning_rate": 2e-05, "loss": 0.0004150390625, "num_tokens": 34669244.0, "reward": 1.0120086669921875, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 31.975, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 35.4, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0125, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.88515625, "reward/mean_r_ndcg": 0.703125, "reward/mean_r_ndcg_cutoff": 5.96875, "reward/mean_recall": 0.2365234375, "reward/mean_recall_at_10": 0.04853515625, "reward/mean_recall_at_100": 0.2365234375, "reward/mean_retrieval_combined_bounded": 0.016683197021484374, "reward/mean_retrieval_combined_raw": 0.016771332174539567, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 576.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.2854574739933014, "rewards/retrieval_reward/mean": 1.0125, "rewards/retrieval_reward/std": 0.2859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.420407295227051, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1189301773905754, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.54387550354004, "sampling/raw_vllm_sampling_logp_difference/mean": 0.32219448685646057, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 580, "step_time": 13.939779492397793 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.35390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.48113387152552606, "epoch": 2.3122529644268774, "frac_reward_zero_std": 0.0, "grad_norm": 2.0073177814483643, "learning_rate": 2e-05, "loss": 0.002783203125, "num_tokens": 34966817.0, "reward": 1.023291015625, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 31.35, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 33.4, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.90625, "reward/mean_r_ndcg": 0.80625, "reward/mean_r_ndcg_cutoff": 6.375, "reward/mean_recall": 0.2640625, "reward/mean_recall_at_10": 0.049169921875, "reward/mean_recall_at_100": 0.2640625, "reward/mean_retrieval_combined_bounded": 0.0318115234375, "reward/mean_retrieval_combined_raw": 0.03690870702266693, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 582.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.2860535174608231, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.28671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7301755428314207, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13456117659807204, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.604998970031739, "sampling/raw_vllm_sampling_logp_difference/mean": 0.31848052740097044, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 585, "step_time": 13.8360214003711 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.31015625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.5109657645225525, "epoch": 2.3320158102766797, "frac_reward_zero_std": 0.0, "grad_norm": 1.6762118339538574, "learning_rate": 2e-05, "loss": -0.001239013671875, "num_tokens": 35265070.0, "reward": 1.0272796630859375, "reward/completion_length_max": 36.4, "reward/completion_length_mean": 31.3, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 33.6, "reward/completion_length_p95": 34.4, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0296875, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.8984375, "reward/mean_r_ndcg": 0.7796875, "reward/mean_r_ndcg_cutoff": 6.01875, "reward/mean_recall": 0.2224609375, "reward/mean_recall_at_10": 0.0490234375, "reward/mean_recall_at_100": 0.2224609375, "reward/mean_retrieval_combined_bounded": 0.027923583984375, "reward/mean_retrieval_combined_raw": 0.04336468279361725, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 586.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2917103886604309, "rewards/retrieval_reward/mean": 1.0296875, "rewards/retrieval_reward/std": 0.291796875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5379591464996336, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13665635883808136, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.505530548095702, "sampling/raw_vllm_sampling_logp_difference/mean": 0.30280232429504395, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 590, "step_time": 13.81538707036525 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.93046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.535495962947607, "epoch": 2.3517786561264824, "frac_reward_zero_std": 0.0, "grad_norm": 1.7057700157165527, "learning_rate": 2e-05, "loss": 0.00027618408203125, "num_tokens": 35563573.0, "reward": 1.0240631103515625, "reward/completion_length_max": 37.2, "reward/completion_length_mean": 31.95, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 34.6, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.87890625, "reward/mean_r_ndcg": 0.75703125, "reward/mean_r_ndcg_cutoff": 5.63125, "reward/mean_recall": 0.22578125, "reward/mean_recall_at_10": 0.0490234375, "reward/mean_recall_at_100": 0.22578125, "reward/mean_retrieval_combined_bounded": 0.0272216796875, "reward/mean_retrieval_combined_raw": 0.021537266671657562, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 592.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28737247586250303, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.287890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4970781564712525, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.11072779148817062, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.97303581237793, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3238803863525391, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 595, "step_time": 13.83677117738407 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.25234375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.2, "completions/min_terminated_length": 0.0, "entropy": 0.5448169231414794, "epoch": 2.3715415019762847, "frac_reward_zero_std": 0.0, "grad_norm": 1.9458622932434082, "learning_rate": 2e-05, "loss": 0.000787353515625, "num_tokens": 35863800.0, "reward": 1.0184234619140624, "reward/completion_length_max": 40.4, "reward/completion_length_mean": 33.3, "reward/completion_length_min": 29.2, "reward/completion_length_p90": 36.2, "reward/completion_length_p95": 37.6, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.0125, "reward/length_out_of_range_rate": 0.01171875, "reward/low_length_query_rate": 0.01171875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.9625, "reward/mean_mrr": 0.9875, "reward/mean_ndcg_at_10": 0.890625, "reward/mean_r_ndcg": 0.7921875, "reward/mean_r_ndcg_cutoff": 6.28125, "reward/mean_recall": 0.217578125, "reward/mean_recall_at_10": 0.048291015625, "reward/mean_recall_at_100": 0.217578125, "reward/mean_retrieval_combined_bounded": 0.036798095703125, "reward/mean_retrieval_combined_raw": 0.06811276525259018, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 597.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01171875, "reward/too_short_query_rate": 0.0, "reward_std": 0.32531079053878786, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.326171875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3066866874694822, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.10066657066345215, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.142879295349122, "sampling/raw_vllm_sampling_logp_difference/mean": 0.32522437572479246, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 600, "step_time": 13.927740937576164 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.36328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.0, "completions/min_terminated_length": 0.0, "entropy": 0.4607121869921684, "epoch": 2.375494071146245, "frac_reward_zero_std": 0.0, "num_tokens": 35924277.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 1.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1137.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6210686357197565, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.6772912941098138, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8751099384344767, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9964819700967458, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6171459682178223, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6706932365236169, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8533333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3267371628660227, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.366171106692776, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5373493975903615, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8554216867469879, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.5902184321252351, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6084627777100807, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.740194957624535, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9184649564050973, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.591340724280876, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.364232578143191, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17850426129778832, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3329237884027527, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 5.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3447.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.404737326830938, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4437172774749511, "retrieval_eval/nq/single_turn/recall_at_10": 0.6471569480707862, "retrieval_eval/nq/single_turn/recall_at_100": 0.8781549173194082, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6655309133756275, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6973611317069189, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8306666666666667, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9386666666666668, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.017425537109375, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 34.25, "reward/completion_length_min": 31.0, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.859375, "reward/mean_r_ndcg": 0.76171875, "reward/mean_r_ndcg_cutoff": 5.25, "reward/mean_recall": 0.228515625, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.228515625, "reward/mean_retrieval_combined_bounded": 0.017333984375, "reward/mean_retrieval_combined_raw": 5.066394805908203e-07, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 600.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2646235227584839, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.265625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.8076006174087524, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.10501302778720856, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.852682113647461, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2932305335998535, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 601, "step_time": 13.897146354895085 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 42.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.31640625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.537669351324439, "epoch": 2.391304347826087, "frac_reward_zero_std": 0.0, "grad_norm": 1.5037362575531006, "learning_rate": 2e-05, "loss": 0.00184326171875, "num_tokens": 36165561.0, "reward": 1.0284233093261719, "reward/completion_length_max": 42.25, "reward/completion_length_mean": 34.3125, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 37.25, "reward/completion_length_p95": 38.5, "reward/duplicate_query_rate": 0.0068359375, "reward/duplicate_ranked_plain_query_rate": 0.0068359375, "reward/invalid_output_rate": 0.0166015625, "reward/length_out_of_range_rate": 0.009765625, "reward/low_length_query_rate": 0.009765625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0302734375, "reward/mean_colbert": 0.9580078125, "reward/mean_mrr": 0.9833984375, "reward/mean_ndcg_at_10": 0.8701171875, "reward/mean_r_ndcg": 0.7734375, "reward/mean_r_ndcg_cutoff": 6.125, "reward/mean_recall": 0.237060546875, "reward/mean_recall_at_10": 0.04803466796875, "reward/mean_recall_at_100": 0.237060546875, "reward/mean_retrieval_combined_bounded": 0.0501708984375, "reward/mean_retrieval_combined_raw": 0.15655517578125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9833984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9833984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 602.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.009765625, "reward/too_short_query_rate": 0.0, "reward_std": 0.31425800919532776, "rewards/retrieval_reward/mean": 1.0302734375, "rewards/retrieval_reward/std": 0.31494140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.167215645313263, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.0903397873044014, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.552433967590332, "sampling/raw_vllm_sampling_logp_difference/mean": 0.32114019244909286, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 605, "step_time": 14.093314963756711 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 43.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.5734375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.4, "completions/min_terminated_length": 0.0, "entropy": 0.5824725709855556, "epoch": 2.411067193675889, "frac_reward_zero_std": 0.0, "grad_norm": 2.264815330505371, "learning_rate": 2e-05, "loss": 0.004150390625, "num_tokens": 36466103.0, "reward": 1.0285980224609375, "reward/completion_length_max": 43.4, "reward/completion_length_mean": 33.55, "reward/completion_length_min": 29.4, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.8, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.01328125, "reward/length_out_of_range_rate": 0.0078125, "reward/low_length_query_rate": 0.0078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.9578125, "reward/mean_mrr": 0.9859375, "reward/mean_ndcg_at_10": 0.86171875, "reward/mean_r_ndcg": 0.66328125, "reward/mean_r_ndcg_cutoff": 6.3125, "reward/mean_recall": 0.221484375, "reward/mean_recall_at_10": 0.047412109375, "reward/mean_recall_at_100": 0.221484375, "reward/mean_retrieval_combined_bounded": 0.045947265625, "reward/mean_retrieval_combined_raw": 0.061670027673244476, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98671875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98671875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 607.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3155051589012146, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.315625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.699676847457886, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.10536318272352219, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.750505256652833, "sampling/raw_vllm_sampling_logp_difference/mean": 0.35671048760414126, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 610, "step_time": 14.243902057188098 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.59140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.4, "completions/min_terminated_length": 0.0, "entropy": 0.561176298558712, "epoch": 2.430830039525692, "frac_reward_zero_std": 0.0, "grad_norm": 1.885324478149414, "learning_rate": 2e-05, "loss": 0.0007354736328125, "num_tokens": 36766700.0, "reward": 1.014276123046875, "reward/completion_length_max": 39.4, "reward/completion_length_mean": 33.55, "reward/completion_length_min": 29.4, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.6, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.00546875, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.98515625, "reward/mean_ndcg_at_10": 0.8671875, "reward/mean_r_ndcg": 0.715625, "reward/mean_r_ndcg_cutoff": 5.85, "reward/mean_recall": 0.2375, "reward/mean_recall_at_10": 0.048388671875, "reward/mean_recall_at_100": 0.2375, "reward/mean_retrieval_combined_bounded": 0.020166015625, "reward/mean_retrieval_combined_raw": 0.031148195266723633, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99453125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99453125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 612.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.29065169394016266, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.2912109375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4143990516662597, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.08688966706395149, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.760269355773925, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3697678089141846, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 615, "step_time": 13.960641240980475 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.43203125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.5196543216705323, "epoch": 2.450592885375494, "frac_reward_zero_std": 0.0, "grad_norm": 1.7719659805297852, "learning_rate": 2e-05, "loss": 0.001080322265625, "num_tokens": 37066869.0, "reward": 0.9956573486328125, "reward/completion_length_max": 41.2, "reward/completion_length_mean": 33.5, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 37.4, "reward/duplicate_query_rate": 0.0109375, "reward/duplicate_ranked_plain_query_rate": 0.0109375, "reward/invalid_output_rate": 0.0328125, "reward/length_out_of_range_rate": 0.021875, "reward/low_length_query_rate": 0.021875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99765625, "reward/mean_colbert": 0.94296875, "reward/mean_mrr": 0.96640625, "reward/mean_ndcg_at_10": 0.865625, "reward/mean_r_ndcg": 0.71796875, "reward/mean_r_ndcg_cutoff": 6.2875, "reward/mean_recall": 0.221484375, "reward/mean_recall_at_10": 0.046826171875, "reward/mean_recall_at_100": 0.221484375, "reward/mean_retrieval_combined_bounded": 0.0394775390625, "reward/mean_retrieval_combined_raw": 0.07060569375753403, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9671875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9671875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 616.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.021875, "reward/too_short_query_rate": 0.0, "reward_std": 0.3559876441955566, "rewards/retrieval_reward/mean": 0.99609375, "rewards/retrieval_reward/std": 0.355859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3949360847473145, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.09391641020774841, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.648305892944336, "sampling/raw_vllm_sampling_logp_difference/mean": 0.34458298683166505, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 620, "step_time": 13.953868827829137 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.4869145542383194, "epoch": 2.4703557312252964, "frac_reward_zero_std": 0.0, "grad_norm": 2.3817033767700195, "learning_rate": 2e-05, "loss": 0.00439453125, "num_tokens": 37366089.0, "reward": 1.0133453369140626, "reward/completion_length_max": 41.8, "reward/completion_length_mean": 32.9, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.01640625, "reward/low_length_query_rate": 0.01640625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01328125, "reward/mean_colbert": 0.95625, "reward/mean_mrr": 0.98046875, "reward/mean_ndcg_at_10": 0.87578125, "reward/mean_r_ndcg": 0.75234375, "reward/mean_r_ndcg_cutoff": 6.1625, "reward/mean_recall": 0.2439453125, "reward/mean_recall_at_10": 0.048291015625, "reward/mean_recall_at_100": 0.2439453125, "reward/mean_retrieval_combined_bounded": 0.0410400390625, "reward/mean_retrieval_combined_raw": 0.06435655578970909, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 622.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01640625, "reward/too_short_query_rate": 0.0, "reward_std": 0.33691610097885133, "rewards/retrieval_reward/mean": 1.01328125, "rewards/retrieval_reward/std": 0.3375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.337021064758301, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.07986343950033188, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.893432807922363, "sampling/raw_vllm_sampling_logp_difference/mean": 0.37026830315589904, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 625, "step_time": 13.994428984075785 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.584375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.5726386047899723, "epoch": 2.4901185770750986, "frac_reward_zero_std": 0.0, "grad_norm": 1.8126449584960938, "learning_rate": 2e-05, "loss": -0.0001678466796875, "num_tokens": 37666709.0, "reward": 1.01524658203125, "reward/completion_length_max": 39.6, "reward/completion_length_mean": 33.55, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 36.6, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.87734375, "reward/mean_r_ndcg": 0.74296875, "reward/mean_r_ndcg_cutoff": 6.3875, "reward/mean_recall": 0.2216796875, "reward/mean_recall_at_10": 0.048486328125, "reward/mean_recall_at_100": 0.2216796875, "reward/mean_retrieval_combined_bounded": 0.0232269287109375, "reward/mean_retrieval_combined_raw": 0.03886733949184418, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 626.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.28372151851654054, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.284375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.443887734413147, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.07505910024046898, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.616697311401367, "sampling/raw_vllm_sampling_logp_difference/mean": 0.36579521298408507, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 630, "step_time": 14.00067857861286 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.6265625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.8, "completions/min_terminated_length": 0.0, "entropy": 0.5920243836939335, "epoch": 2.5098814229249014, "frac_reward_zero_std": 0.0, "grad_norm": 1.5123381614685059, "learning_rate": 2e-05, "loss": 0.0032958984375, "num_tokens": 37968567.0, "reward": 1.0243804931640625, "reward/completion_length_max": 40.8, "reward/completion_length_mean": 34.65, "reward/completion_length_min": 30.8, "reward/completion_length_p90": 36.8, "reward/completion_length_p95": 37.6, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.0140625, "reward/length_out_of_range_rate": 0.00625, "reward/low_length_query_rate": 0.00625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.96015625, "reward/mean_mrr": 0.9859375, "reward/mean_ndcg_at_10": 0.89609375, "reward/mean_r_ndcg": 0.74765625, "reward/mean_r_ndcg_cutoff": 6.3375, "reward/mean_recall": 0.2306640625, "reward/mean_recall_at_10": 0.047802734375, "reward/mean_recall_at_100": 0.2306640625, "reward/mean_retrieval_combined_bounded": 0.041455078125, "reward/mean_retrieval_combined_raw": 0.07118982076644897, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9859375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9859375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 632.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3058332145214081, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.306640625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6815916538238525, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.09382817149162292, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.840413284301758, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3299198389053345, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 635, "step_time": 14.074534257792402 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 42.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.459375, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.8, "completions/min_terminated_length": 0.0, "entropy": 0.5920812383294105, "epoch": 2.5296442687747036, "frac_reward_zero_std": 0.0, "grad_norm": 2.1455562114715576, "learning_rate": 2e-05, "loss": 0.00693359375, "num_tokens": 38272995.0, "reward": 1.02230224609375, "reward/completion_length_max": 42.8, "reward/completion_length_mean": 36.45, "reward/completion_length_min": 31.8, "reward/completion_length_p90": 39.0, "reward/completion_length_p95": 39.8, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.02578125, "reward/length_out_of_range_rate": 0.021875, "reward/low_length_query_rate": 0.021875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.95078125, "reward/mean_mrr": 0.97421875, "reward/mean_ndcg_at_10": 0.88125, "reward/mean_r_ndcg": 0.74453125, "reward/mean_r_ndcg_cutoff": 5.89375, "reward/mean_recall": 0.2197265625, "reward/mean_recall_at_10": 0.047607421875, "reward/mean_recall_at_100": 0.2197265625, "reward/mean_retrieval_combined_bounded": 0.0588623046875, "reward/mean_retrieval_combined_raw": 0.1115234375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97421875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97421875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 637.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.021875, "reward/too_short_query_rate": 0.0, "reward_std": 0.3498194992542267, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.35078125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.1923458576202393, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.06548608094453812, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.036189079284668, "sampling/raw_vllm_sampling_logp_difference/mean": 0.35222278237342836, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 640, "step_time": 14.140199318004306 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 43.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.70625, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.0, "completions/min_terminated_length": 0.0, "entropy": 0.5381974898278713, "epoch": 2.549407114624506, "frac_reward_zero_std": 0.0, "grad_norm": 1.6355812549591064, "learning_rate": 2e-05, "loss": 0.00849609375, "num_tokens": 38577643.0, "reward": 0.95113525390625, "reward/completion_length_max": 43.6, "reward/completion_length_mean": 36.7, "reward/completion_length_min": 32.0, "reward/completion_length_p90": 40.0, "reward/completion_length_p95": 40.6, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.08046875, "reward/length_out_of_range_rate": 0.07265625, "reward/low_length_query_rate": 0.07265625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.94921875, "reward/mean_colbert": 0.89609375, "reward/mean_mrr": 0.91953125, "reward/mean_ndcg_at_10": 0.83359375, "reward/mean_r_ndcg": 0.7578125, "reward/mean_r_ndcg_cutoff": 5.725, "reward/mean_recall": 0.211328125, "reward/mean_recall_at_10": 0.044384765625, "reward/mean_recall_at_100": 0.211328125, "reward/mean_retrieval_combined_bounded": 0.067919921875, "reward/mean_retrieval_combined_raw": 0.160546875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.91953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.91953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 641.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.07265625, "reward/too_short_query_rate": 0.0, "reward_std": 0.43939064145088197, "rewards/retrieval_reward/mean": 0.94921875, "rewards/retrieval_reward/std": 0.440625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4735183715820312, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.10269702076911927, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.595357513427734, "sampling/raw_vllm_sampling_logp_difference/mean": 0.33243429064750674, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 645, "step_time": 14.114367963990663 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.95234375, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.0, "completions/min_terminated_length": 0.0, "entropy": 0.5788900889456272, "epoch": 2.5691699604743086, "frac_reward_zero_std": 0.0, "grad_norm": 1.9307658672332764, "learning_rate": 2e-05, "loss": 0.00283203125, "num_tokens": 38878478.0, "reward": 1.0111846923828125, "reward/completion_length_max": 41.8, "reward/completion_length_mean": 34.0, "reward/completion_length_min": 30.0, "reward/completion_length_p90": 36.4, "reward/completion_length_p95": 37.2, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.00859375, "reward/length_out_of_range_rate": 0.00703125, "reward/low_length_query_rate": 0.00703125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01171875, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.99140625, "reward/mean_ndcg_at_10": 0.87890625, "reward/mean_r_ndcg": 0.75703125, "reward/mean_r_ndcg_cutoff": 5.94375, "reward/mean_recall": 0.221484375, "reward/mean_recall_at_10": 0.047412109375, "reward/mean_recall_at_100": 0.221484375, "reward/mean_retrieval_combined_bounded": 0.023291015625, "reward/mean_retrieval_combined_raw": 0.034375869855284694, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99140625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99140625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 647.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00703125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3048424482345581, "rewards/retrieval_reward/mean": 1.01171875, "rewards/retrieval_reward/std": 0.30546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.358784055709839, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.08242426291108132, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.177977752685546, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3240475058555603, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 650, "step_time": 14.045866650168318 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.9453125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.5326741635799408, "epoch": 2.5731225296442686, "frac_reward_zero_std": 0.0, "num_tokens": 38938048.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6278482769160034, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.6844231689879223, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8813708260105448, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9956063268892794, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6244676688549899, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6774726145464316, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.86, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3318503682540311, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.37052079972074864, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5349397590361445, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8385542168674699, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.5850878330732594, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6025343476692905, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7462662337662337, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.8998056520943846, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5874089216913205, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.363001158286898, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.1780404797667715, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3207387256242786, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 1.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3451.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.40331695374966714, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.44084426103099833, "retrieval_eval/nq/single_turn/recall_at_10": 0.6421568627450981, "retrieval_eval/nq/single_turn/recall_at_100": 0.8745291219936251, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6556664250662789, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6893454366543114, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8233333333333334, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9346666666666666, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 0.9791412353515625, "reward/completion_length_max": 40.0, "reward/completion_length_mean": 32.0, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.98046875, "reward/mean_colbert": 0.9765625, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.85546875, "reward/mean_r_ndcg": 0.76171875, "reward/mean_r_ndcg_cutoff": 4.375, "reward/mean_recall": 0.21875, "reward/mean_recall_at_10": 0.049560546875, "reward/mean_recall_at_100": 0.21875, "reward/mean_retrieval_combined_bounded": -0.0206298828125, "reward/mean_retrieval_combined_raw": -0.00014019012451171875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 648.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2699022591114044, "rewards/retrieval_reward/mean": 0.98046875, "rewards/retrieval_reward/std": 0.271484375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8977339267730713, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.0944988951086998, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.191083908081055, "sampling/raw_vllm_sampling_logp_difference/mean": 0.34457385540008545, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 651, "step_time": 13.902017108921427 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.7509765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.25, "completions/min_terminated_length": 0.0, "entropy": 0.4846226079389453, "epoch": 2.588932806324111, "frac_reward_zero_std": 0.0, "grad_norm": 1.7645504474639893, "learning_rate": 2e-05, "loss": -0.00255126953125, "num_tokens": 39176289.0, "reward": 1.0061416625976562, "reward/completion_length_max": 36.0, "reward/completion_length_mean": 31.75, "reward/completion_length_min": 28.25, "reward/completion_length_p90": 33.5, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.001953125, "reward/duplicate_ranked_plain_query_rate": 0.001953125, "reward/invalid_output_rate": 0.001953125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0048828125, "reward/mean_colbert": 0.9736328125, "reward/mean_mrr": 0.998046875, "reward/mean_ndcg_at_10": 0.896484375, "reward/mean_r_ndcg": 0.7783203125, "reward/mean_r_ndcg_cutoff": 5.8984375, "reward/mean_recall": 0.221923828125, "reward/mean_recall_at_10": 0.04815673828125, "reward/mean_recall_at_100": 0.221923828125, "reward/mean_retrieval_combined_bounded": 0.008051872253417969, "reward/mean_retrieval_combined_raw": 0.0134104760363698, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.998046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.998046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 653.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.299118272960186, "rewards/retrieval_reward/mean": 1.0048828125, "rewards/retrieval_reward/std": 0.30029296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.662236452102661, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13274919986724854, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.808041095733643, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2876540422439575, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 655, "step_time": 13.813521679199766 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.246875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.4, "completions/min_terminated_length": 0.0, "entropy": 0.45006473213434217, "epoch": 2.608695652173913, "frac_reward_zero_std": 0.0, "grad_norm": 2.4251787662506104, "learning_rate": 2e-05, "loss": -0.00012969970703125, "num_tokens": 39474109.0, "reward": 1.0055389404296875, "reward/completion_length_max": 35.6, "reward/completion_length_mean": 31.225, "reward/completion_length_min": 27.4, "reward/completion_length_p90": 33.8, "reward/completion_length_p95": 34.4, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0046875, "reward/mean_colbert": 0.975, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.91484375, "reward/mean_r_ndcg": 0.76171875, "reward/mean_r_ndcg_cutoff": 6.8, "reward/mean_recall": 0.2359375, "reward/mean_recall_at_10": 0.04931640625, "reward/mean_recall_at_100": 0.2359375, "reward/mean_retrieval_combined_bounded": 0.0054443359375, "reward/mean_retrieval_combined_raw": 1.6771256923675537e-06, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 656.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2867615044116974, "rewards/retrieval_reward/mean": 1.0046875, "rewards/retrieval_reward/std": 0.2875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4782801628112794, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12047789096832276, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.59546184539795, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3141523003578186, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 660, "step_time": 13.780050955410115 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.5171875, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.6, "completions/min_terminated_length": 0.0, "entropy": 0.48824648931622505, "epoch": 2.6284584980237153, "frac_reward_zero_std": 0.0, "grad_norm": 1.6806321144104004, "learning_rate": 2e-05, "loss": -0.0018310546875, "num_tokens": 39770803.0, "reward": 1.012493896484375, "reward/completion_length_max": 35.6, "reward/completion_length_mean": 30.475, "reward/completion_length_min": 26.6, "reward/completion_length_p90": 32.4, "reward/completion_length_p95": 33.2, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.97421875, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.89453125, "reward/mean_r_ndcg": 0.759375, "reward/mean_r_ndcg_cutoff": 6.2375, "reward/mean_recall": 0.2318359375, "reward/mean_recall_at_10": 0.04970703125, "reward/mean_recall_at_100": 0.2318359375, "reward/mean_retrieval_combined_bounded": 0.0139984130859375, "reward/mean_retrieval_combined_raw": 0.008898922801017761, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 662.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2892254114151001, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.289453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7135294914245605, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1283695936203003, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.297270965576171, "sampling/raw_vllm_sampling_logp_difference/mean": 0.31403645873069763, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 665, "step_time": 13.752484921179711 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.78984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.2, "completions/min_terminated_length": 0.0, "entropy": 0.43005336150527, "epoch": 2.6482213438735176, "frac_reward_zero_std": 0.0, "grad_norm": 1.8754208087921143, "learning_rate": 2e-05, "loss": -0.00262451171875, "num_tokens": 40067302.0, "reward": 1.0076263427734375, "reward/completion_length_max": 36.6, "reward/completion_length_mean": 30.8, "reward/completion_length_min": 27.2, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00703125, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.99765625, "reward/mean_ndcg_at_10": 0.91015625, "reward/mean_r_ndcg": 0.7484375, "reward/mean_r_ndcg_cutoff": 6.18125, "reward/mean_recall": 0.246484375, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.246484375, "reward/mean_retrieval_combined_bounded": 0.0100128173828125, "reward/mean_retrieval_combined_raw": 0.01899574287235737, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 666.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2893402576446533, "rewards/retrieval_reward/mean": 1.00703125, "rewards/retrieval_reward/std": 0.28984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6368866920471192, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12238912433385848, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.098052978515625, "sampling/raw_vllm_sampling_logp_difference/mean": 0.28501731157302856, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 670, "step_time": 13.778601876820904 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.215625, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.45882013663649557, "epoch": 2.6679841897233203, "frac_reward_zero_std": 0.0, "grad_norm": 2.2111756801605225, "learning_rate": 2e-05, "loss": -0.00135498046875, "num_tokens": 40363322.0, "reward": 0.99970703125, "reward/completion_length_max": 34.8, "reward/completion_length_mean": 30.225, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.01796875, "reward/duplicate_ranked_plain_query_rate": 0.01796875, "reward/invalid_output_rate": 0.01796875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0, "reward/mean_colbert": 0.9578125, "reward/mean_mrr": 0.975, "reward/mean_ndcg_at_10": 0.8546875, "reward/mean_r_ndcg": 0.74375, "reward/mean_r_ndcg_cutoff": 6.7125, "reward/mean_recall": 0.225, "reward/mean_recall_at_10": 0.0470703125, "reward/mean_recall_at_100": 0.225, "reward/mean_retrieval_combined_bounded": 0.017766571044921874, "reward/mean_retrieval_combined_raw": 0.016506360471248628, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98203125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98203125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 672.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.30053259134292604, "rewards/retrieval_reward/mean": 1.0, "rewards/retrieval_reward/std": 0.30078125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.857761573791504, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12688596099615096, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.406108379364014, "sampling/raw_vllm_sampling_logp_difference/mean": 0.31270393133163454, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 675, "step_time": 13.624905292189215 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.103125, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.4, "completions/min_terminated_length": 0.0, "entropy": 0.46447652354836466, "epoch": 2.6877470355731226, "frac_reward_zero_std": 0.0, "grad_norm": 2.196446657180786, "learning_rate": 2e-05, "loss": -0.001202392578125, "num_tokens": 40659230.0, "reward": 1.016961669921875, "reward/completion_length_max": 35.2, "reward/completion_length_mean": 30.1, "reward/completion_length_min": 26.4, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 32.8, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.975, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.9078125, "reward/mean_r_ndcg": 0.7765625, "reward/mean_r_ndcg_cutoff": 6.01875, "reward/mean_recall": 0.23828125, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.23828125, "reward/mean_retrieval_combined_bounded": 0.018511962890625, "reward/mean_retrieval_combined_raw": 0.006841894611716271, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 677.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2768613278865814, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.276953125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6617247581481935, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12698912620544434, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.169682884216309, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3210075259208679, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 680, "step_time": 13.680558285256847 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 53.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.11171875, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.2, "completions/min_terminated_length": 0.0, "entropy": 0.45993765592575075, "epoch": 2.707509881422925, "frac_reward_zero_std": 0.0, "grad_norm": 2.094050168991089, "learning_rate": 2e-05, "loss": 0.01611328125, "num_tokens": 40955341.0, "reward": 1.0239166259765624, "reward/completion_length_max": 53.2, "reward/completion_length_mean": 30.1, "reward/completion_length_min": 26.2, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 32.6, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0015625, "reward/malformed_ranked_plain_query_rate": 0.0015625, "reward/mean": 1.025, "reward/mean_colbert": 0.9703125, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.8640625, "reward/mean_r_ndcg": 0.76484375, "reward/mean_r_ndcg_cutoff": 5.25, "reward/mean_recall": 0.2185546875, "reward/mean_recall_at_10": 0.047705078125, "reward/mean_recall_at_100": 0.2185546875, "reward/mean_retrieval_combined_bounded": 0.0325927734375, "reward/mean_retrieval_combined_raw": 0.06080479989759624, "reward/output_tag_rate": 0.9984375, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 681.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3157669186592102, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.316015625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7493980884552003, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16371484994888305, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.626044464111328, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26943387985229494, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 685, "step_time": 14.569661467405968 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.09375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.46801728680729865, "epoch": 2.7272727272727275, "frac_reward_zero_std": 0.0, "grad_norm": 1.7861626148223877, "learning_rate": 2e-05, "loss": -0.0007415771484375, "num_tokens": 41252965.0, "reward": 1.0191253662109374, "reward/completion_length_max": 36.4, "reward/completion_length_mean": 31.075, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 33.4, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01953125, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.85859375, "reward/mean_r_ndcg": 0.68984375, "reward/mean_r_ndcg_cutoff": 5.95625, "reward/mean_recall": 0.2287109375, "reward/mean_recall_at_10": 0.048779296875, "reward/mean_recall_at_100": 0.2287109375, "reward/mean_retrieval_combined_bounded": 0.0223480224609375, "reward/mean_retrieval_combined_raw": 0.016993305087089537, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 687.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28019399642944337, "rewards/retrieval_reward/mean": 1.01953125, "rewards/retrieval_reward/std": 0.28046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5738444566726684, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12561038881540298, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.57185230255127, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3100397914648056, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 690, "step_time": 13.798688149405644 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.45625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.5174344807863236, "epoch": 2.7470355731225298, "frac_reward_zero_std": 0.0, "grad_norm": 2.22125506401062, "learning_rate": 2e-05, "loss": 0.0052001953125, "num_tokens": 41553133.0, "reward": 1.0075286865234374, "reward/completion_length_max": 41.6, "reward/completion_length_mean": 33.4, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 36.2, "reward/completion_length_p95": 37.6, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.02109375, "reward/length_out_of_range_rate": 0.015625, "reward/low_length_query_rate": 0.015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00703125, "reward/mean_colbert": 0.95703125, "reward/mean_mrr": 0.97890625, "reward/mean_ndcg_at_10": 0.88125, "reward/mean_r_ndcg": 0.78984375, "reward/mean_r_ndcg_cutoff": 6.1625, "reward/mean_recall": 0.225390625, "reward/mean_recall_at_10": 0.047900390625, "reward/mean_recall_at_100": 0.225390625, "reward/mean_retrieval_combined_bounded": 0.036398696899414065, "reward/mean_retrieval_combined_raw": 0.06957770586013794, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 692.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.32165075540542604, "rewards/retrieval_reward/mean": 1.00703125, "rewards/retrieval_reward/std": 0.321875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7187079906463625, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12005644291639328, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.977120590209961, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3064598679542542, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 695, "step_time": 14.042585462448187 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.60390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.4, "completions/min_terminated_length": 0.0, "entropy": 0.5551575817167759, "epoch": 2.766798418972332, "frac_reward_zero_std": 0.0, "grad_norm": 1.523349642753601, "learning_rate": 2e-05, "loss": 0.003662109375, "num_tokens": 41854866.0, "reward": 1.0078338623046874, "reward/completion_length_max": 40.8, "reward/completion_length_mean": 34.6, "reward/completion_length_min": 30.4, "reward/completion_length_p90": 37.2, "reward/completion_length_p95": 38.0, "reward/duplicate_query_rate": 0.03046875, "reward/duplicate_ranked_plain_query_rate": 0.03046875, "reward/invalid_output_rate": 0.0390625, "reward/length_out_of_range_rate": 0.00859375, "reward/low_length_query_rate": 0.00859375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00859375, "reward/mean_colbert": 0.93671875, "reward/mean_mrr": 0.9609375, "reward/mean_ndcg_at_10": 0.84609375, "reward/mean_r_ndcg": 0.715625, "reward/mean_r_ndcg_cutoff": 6.5375, "reward/mean_recall": 0.1830078125, "reward/mean_recall_at_10": 0.04619140625, "reward/mean_recall_at_100": 0.1830078125, "reward/mean_retrieval_combined_bounded": 0.051251220703125, "reward/mean_retrieval_combined_raw": 0.10252823829650878, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 696.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00859375, "reward/too_short_query_rate": 0.0, "reward_std": 0.34710089564323426, "rewards/retrieval_reward/mean": 1.00859375, "rewards/retrieval_reward/std": 0.347265625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7457135677337647, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.10467294901609421, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.138332557678222, "sampling/raw_vllm_sampling_logp_difference/mean": 0.32016443014144896, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 700, "step_time": 13.990001859969926 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 44.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.8828125, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.0, "completions/min_terminated_length": 0.0, "entropy": 0.5536227822303772, "epoch": 2.7707509881422925, "frac_reward_zero_std": 0.0, "num_tokens": 41915284.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6446876801625449, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.698526333638741, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8848857644991213, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9947275922671354, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5811862842575549, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6306889701488976, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8066666666666666, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.33250105781383654, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3621551467612001, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5012048192771085, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8313253012048193, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6174129319150582, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6118848866541821, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7264374123529054, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.8942907444668008, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5764134693723574, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.35723063506469227, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.1729075777259061, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.30848943265677026, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 4.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3448.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.40700181492976995, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.44274834019867404, "retrieval_eval/nq/single_turn/recall_at_10": 0.6404437354988399, "retrieval_eval/nq/single_turn/recall_at_100": 0.8761600928074246, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6582951455936094, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6922474162720694, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8328888888888889, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9510000000000001, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.018951416015625, "reward/completion_length_max": 44.0, "reward/completion_length_mean": 35.0, "reward/completion_length_min": 31.0, "reward/completion_length_p90": 39.0, "reward/completion_length_p95": 40.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.01953125, "reward/low_length_query_rate": 0.01953125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.95703125, "reward/mean_mrr": 0.9765625, "reward/mean_ndcg_at_10": 0.8359375, "reward/mean_r_ndcg": 0.71484375, "reward/mean_r_ndcg_cutoff": 8.3125, "reward/mean_recall": 0.2333984375, "reward/mean_recall_at_10": 0.044921875, "reward/mean_recall_at_100": 0.2333984375, "reward/mean_retrieval_combined_bounded": 0.048095703125, "reward/mean_retrieval_combined_raw": 0.0927734375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 700.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01953125, "reward/too_short_query_rate": 0.0, "reward_std": 0.32956692576408386, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.330078125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.3284616470336914, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.04707159847021103, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.357852935791016, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3674735128879547, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 701, "step_time": 14.15338721103035 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.537109375, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.0, "completions/min_terminated_length": 0.0, "entropy": 0.5574926026165485, "epoch": 2.7865612648221343, "frac_reward_zero_std": 0.0, "grad_norm": 1.807532787322998, "learning_rate": 2e-05, "loss": 0.00230712890625, "num_tokens": 42156474.0, "reward": 1.0266151428222656, "reward/completion_length_max": 40.0, "reward/completion_length_mean": 34.5625, "reward/completion_length_min": 30.0, "reward/completion_length_p90": 36.75, "reward/completion_length_p95": 37.5, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.0068359375, "reward/length_out_of_range_rate": 0.0029296875, "reward/low_length_query_rate": 0.0029296875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025390625, "reward/mean_colbert": 0.9716796875, "reward/mean_mrr": 0.9931640625, "reward/mean_ndcg_at_10": 0.8486328125, "reward/mean_r_ndcg": 0.724609375, "reward/mean_r_ndcg_cutoff": 5.7734375, "reward/mean_recall": 0.224853515625, "reward/mean_recall_at_10": 0.04876708984375, "reward/mean_recall_at_100": 0.224853515625, "reward/mean_retrieval_combined_bounded": 0.03497314453125, "reward/mean_retrieval_combined_raw": 0.04864501953125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9931640625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9931640625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 703.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0029296875, "reward/too_short_query_rate": 0.0, "reward_std": 0.2873404249548912, "rewards/retrieval_reward/mean": 1.025390625, "rewards/retrieval_reward/std": 0.28759765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.439471423625946, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.08082645758986473, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.991600275039673, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3503371328115463, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 705, "step_time": 14.052453023192356 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.340625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.6, "completions/min_terminated_length": 0.0, "entropy": 0.5275723151862621, "epoch": 2.8063241106719365, "frac_reward_zero_std": 0.0, "grad_norm": 1.7203727960586548, "learning_rate": 2e-05, "loss": 0.000970458984375, "num_tokens": 42457742.0, "reward": 1.01129150390625, "reward/completion_length_max": 40.4, "reward/completion_length_mean": 34.4, "reward/completion_length_min": 29.6, "reward/completion_length_p90": 37.0, "reward/completion_length_p95": 38.2, "reward/duplicate_query_rate": 0.00859375, "reward/duplicate_ranked_plain_query_rate": 0.00859375, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.0109375, "reward/low_length_query_rate": 0.0109375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.009375, "reward/mean_colbert": 0.95859375, "reward/mean_mrr": 0.98046875, "reward/mean_ndcg_at_10": 0.90546875, "reward/mean_r_ndcg": 0.8, "reward/mean_r_ndcg_cutoff": 6.8, "reward/mean_recall": 0.2203125, "reward/mean_recall_at_10": 0.047998046875, "reward/mean_recall_at_100": 0.2203125, "reward/mean_retrieval_combined_bounded": 0.03619384765625, "reward/mean_retrieval_combined_raw": 0.0583212673664093, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 706.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0109375, "reward/too_short_query_rate": 0.0, "reward_std": 0.3212704062461853, "rewards/retrieval_reward/mean": 1.009375, "rewards/retrieval_reward/std": 0.321484375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3542521238327025, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.09754644781351089, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.267683029174805, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3436054527759552, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 710, "step_time": 14.041709816385993 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.7703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.2, "completions/min_terminated_length": 0.0, "entropy": 0.5063788592815399, "epoch": 2.8260869565217392, "frac_reward_zero_std": 0.0, "grad_norm": 2.514570951461792, "learning_rate": 2e-05, "loss": 0.001507568359375, "num_tokens": 42758632.0, "reward": 1.0279144287109374, "reward/completion_length_max": 40.4, "reward/completion_length_mean": 33.75, "reward/completion_length_min": 29.2, "reward/completion_length_p90": 37.4, "reward/completion_length_p95": 38.0, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.003125, "reward/low_length_query_rate": 0.003125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.8828125, "reward/mean_r_ndcg": 0.803125, "reward/mean_r_ndcg_cutoff": 6.2, "reward/mean_recall": 0.2498046875, "reward/mean_recall_at_10": 0.049169921875, "reward/mean_recall_at_100": 0.2498046875, "reward/mean_retrieval_combined_bounded": 0.03360595703125, "reward/mean_retrieval_combined_raw": 0.018954092264175416, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 712.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.003125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3014044165611267, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.30234375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.9675284624099731, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.060637868195772174, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 13.857499694824218, "sampling/raw_vllm_sampling_logp_difference/mean": 0.41012628078460694, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 715, "step_time": 14.044897434243467 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 55.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.31640625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.47917484417557715, "epoch": 2.8458498023715415, "frac_reward_zero_std": 0.0, "grad_norm": 2.2519168853759766, "learning_rate": 2e-05, "loss": 0.01279296875, "num_tokens": 43057053.0, "reward": 1.0267852783203124, "reward/completion_length_max": 55.4, "reward/completion_length_mean": 32.3, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.028125, "reward/mean_colbert": 0.9765625, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.8875, "reward/mean_r_ndcg": 0.765625, "reward/mean_r_ndcg_cutoff": 6.36875, "reward/mean_recall": 0.2166015625, "reward/mean_recall_at_10": 0.048876953125, "reward/mean_recall_at_100": 0.2166015625, "reward/mean_retrieval_combined_bounded": 0.0291259765625, "reward/mean_retrieval_combined_raw": 0.006348365545272827, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 717.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3002969861030579, "rewards/retrieval_reward/mean": 1.028125, "rewards/retrieval_reward/std": 0.30078125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.2303179025650026, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.06288176476955414, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.443243789672852, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3830512225627899, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 720, "step_time": 14.663414282805752 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.65078125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.43185103833675387, "epoch": 2.8656126482213438, "frac_reward_zero_std": 0.0, "grad_norm": 1.8722517490386963, "learning_rate": 2e-05, "loss": -0.0031005859375, "num_tokens": 43354814.0, "reward": 1.0153076171875, "reward/completion_length_max": 36.0, "reward/completion_length_mean": 31.625, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 34.6, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01484375, "reward/mean_colbert": 0.97421875, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.90390625, "reward/mean_r_ndcg": 0.73125, "reward/mean_r_ndcg_cutoff": 6.41875, "reward/mean_recall": 0.233984375, "reward/mean_recall_at_10": 0.048974609375, "reward/mean_recall_at_100": 0.233984375, "reward/mean_retrieval_combined_bounded": 0.0162200927734375, "reward/mean_retrieval_combined_raw": 0.006444378942251206, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 721.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2971168875694275, "rewards/retrieval_reward/mean": 1.01484375, "rewards/retrieval_reward/std": 0.29765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.1892683029174806, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.0922737568616867, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.161867141723633, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3812728226184845, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 725, "step_time": 13.774702332192101 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.69296875, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.2, "completions/min_terminated_length": 0.0, "entropy": 0.45814650505781174, "epoch": 2.8853754940711465, "frac_reward_zero_std": 0.0, "grad_norm": 2.5227487087249756, "learning_rate": 2e-05, "loss": 0.00037689208984375, "num_tokens": 43651989.0, "reward": 1.001531982421875, "reward/completion_length_max": 35.4, "reward/completion_length_mean": 30.75, "reward/completion_length_min": 26.2, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 33.6, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00078125, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.85546875, "reward/mean_r_ndcg": 0.778125, "reward/mean_r_ndcg_cutoff": 5.59375, "reward/mean_recall": 0.2115234375, "reward/mean_recall_at_10": 0.046826171875, "reward/mean_recall_at_100": 0.2115234375, "reward/mean_retrieval_combined_bounded": 0.0079620361328125, "reward/mean_retrieval_combined_raw": 0.02144017517566681, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 727.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.283872252702713, "rewards/retrieval_reward/mean": 1.00078125, "rewards/retrieval_reward/std": 0.284375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.106666374206543, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.06716154739260674, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.88147201538086, "sampling/raw_vllm_sampling_logp_difference/mean": 0.40418183207511904, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 730, "step_time": 13.739098392613233 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.1984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.2, "completions/min_terminated_length": 0.0, "entropy": 0.4519004873931408, "epoch": 2.9051383399209487, "frac_reward_zero_std": 0.0, "grad_norm": 2.221970558166504, "learning_rate": 2e-05, "loss": -0.001019287109375, "num_tokens": 43947955.0, "reward": 1.004290771484375, "reward/completion_length_max": 35.8, "reward/completion_length_mean": 30.2, "reward/completion_length_min": 26.2, "reward/completion_length_p90": 32.4, "reward/completion_length_p95": 33.6, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0046875, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.84453125, "reward/mean_r_ndcg": 0.721875, "reward/mean_r_ndcg_cutoff": 6.01875, "reward/mean_recall": 0.240625, "reward/mean_recall_at_10": 0.047802734375, "reward/mean_recall_at_100": 0.240625, "reward/mean_retrieval_combined_bounded": 0.00751953125, "reward/mean_retrieval_combined_raw": 0.0098267771769315, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 732.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28277690410614015, "rewards/retrieval_reward/mean": 1.003125, "rewards/retrieval_reward/std": 0.283203125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.10937876701355, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.06861600577831269, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.329999351501465, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3728539526462555, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 735, "step_time": 13.719905847439076 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.9375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.46672193259000777, "epoch": 2.924901185770751, "frac_reward_zero_std": 0.0, "grad_norm": 2.451214075088501, "learning_rate": 2e-05, "loss": -0.0010986328125, "num_tokens": 44246467.0, "reward": 1.0112518310546874, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 31.95, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 35.4, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01171875, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.85546875, "reward/mean_r_ndcg": 0.725, "reward/mean_r_ndcg_cutoff": 6.59375, "reward/mean_recall": 0.2546875, "reward/mean_recall_at_10": 0.04921875, "reward/mean_recall_at_100": 0.2546875, "reward/mean_retrieval_combined_bounded": 0.012068939208984376, "reward/mean_retrieval_combined_raw": 0.006073915585875511, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 736.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2934859454631805, "rewards/retrieval_reward/mean": 1.01171875, "rewards/retrieval_reward/std": 0.294140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.275969219207764, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.05991761535406113, "sampling/raw_vllm_importance_sampling_ratio/min": 1.2402731619509667e-12, "sampling/raw_vllm_sampling_logp_difference/max": 11.485723876953125, "sampling/raw_vllm_sampling_logp_difference/mean": 0.39903382658958436, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 740, "step_time": 13.840459173801355 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.24140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.4696327894926071, "epoch": 2.9446640316205532, "frac_reward_zero_std": 0.0, "grad_norm": 2.0679852962493896, "learning_rate": 2e-05, "loss": -0.00024871826171875, "num_tokens": 44545048.0, "reward": 1.0305511474609375, "reward/completion_length_max": 37.4, "reward/completion_length_mean": 32.225, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.975, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.92734375, "reward/mean_r_ndcg": 0.796875, "reward/mean_r_ndcg_cutoff": 6.85, "reward/mean_recall": 0.22109375, "reward/mean_recall_at_10": 0.048974609375, "reward/mean_recall_at_100": 0.22109375, "reward/mean_retrieval_combined_bounded": 0.0319091796875, "reward/mean_retrieval_combined_raw": 0.05009777843952179, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 742.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27542670965194704, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.27578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.272959756851196, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.0684853620827198, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.278964614868164, "sampling/raw_vllm_sampling_logp_difference/mean": 0.38310173749923704, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 745, "step_time": 13.851717048173304 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.21796875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.8, "completions/min_terminated_length": 0.0, "entropy": 0.46883343532681465, "epoch": 2.9644268774703555, "frac_reward_zero_std": 0.0, "grad_norm": 2.0814733505249023, "learning_rate": 2e-05, "loss": 0.0006927490234375, "num_tokens": 44844879.0, "reward": 1.013873291015625, "reward/completion_length_max": 39.6, "reward/completion_length_mean": 33.25, "reward/completion_length_min": 28.8, "reward/completion_length_p90": 36.2, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01484375, "reward/mean_colbert": 0.9703125, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.8890625, "reward/mean_r_ndcg": 0.728125, "reward/mean_r_ndcg_cutoff": 6.14375, "reward/mean_recall": 0.2171875, "reward/mean_recall_at_10": 0.048291015625, "reward/mean_recall_at_100": 0.2171875, "reward/mean_retrieval_combined_bounded": 0.0189605712890625, "reward/mean_retrieval_combined_raw": 0.01971386671066284, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 746.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2807815849781036, "rewards/retrieval_reward/mean": 1.01484375, "rewards/retrieval_reward/std": 0.280859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.193021297454834, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.059923330694437026, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.217131805419921, "sampling/raw_vllm_sampling_logp_difference/mean": 0.41774205565452577, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 750, "step_time": 14.024118066008668 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.12890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.0, "completions/min_terminated_length": 0.0, "entropy": 0.5543155819177628, "epoch": 2.9683794466403164, "frac_reward_zero_std": 0.0, "num_tokens": 44904880.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6543644071189111, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7073867218261665, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8892794376098418, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9956063268892794, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6112502698659168, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6541047436459417, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8066666666666666, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.318299938637495, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3590685969974799, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5277108433734939, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8289156626506025, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6101552887795536, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6207631309067007, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7473492469971343, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9014377934272301, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5666536673670787, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3544641506767641, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17904681027302072, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3063772946553482, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 4.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3448.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.4102828974507484, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4450733041153394, "retrieval_eval/nq/single_turn/recall_at_10": 0.6395253286929622, "retrieval_eval/nq/single_turn/recall_at_100": 0.8755800464037123, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6817457613515364, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7073313330911412, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8248333333333333, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9460000000000001, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.024810791015625, "reward/completion_length_max": 39.0, "reward/completion_length_mean": 33.25, "reward/completion_length_min": 30.0, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.76953125, "reward/mean_r_ndcg": 0.55859375, "reward/mean_r_ndcg_cutoff": 5.84375, "reward/mean_recall": 0.232421875, "reward/mean_recall_at_10": 0.0458984375, "reward/mean_recall_at_100": 0.232421875, "reward/mean_retrieval_combined_bounded": 0.0286865234375, "reward/mean_retrieval_combined_raw": 0.031982421875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 752.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.282399982213974, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.283203125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.9847280979156494, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.058733586221933365, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 15.728738784790039, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4303196668624878, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 751, "step_time": 13.979414327070117 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.5126953125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.75, "completions/min_terminated_length": 0.0, "entropy": 0.4701820146292448, "epoch": 2.984189723320158, "frac_reward_zero_std": 0.0, "grad_norm": 2.011831283569336, "learning_rate": 2e-05, "loss": 0.0002899169921875, "num_tokens": 45145245.0, "reward": 1.030731201171875, "reward/completion_length_max": 39.25, "reward/completion_length_mean": 33.5625, "reward/completion_length_min": 28.75, "reward/completion_length_p90": 36.25, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.0087890625, "reward/duplicate_ranked_plain_query_rate": 0.0087890625, "reward/invalid_output_rate": 0.0107421875, "reward/length_out_of_range_rate": 0.001953125, "reward/low_length_query_rate": 0.001953125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.9892578125, "reward/mean_ndcg_at_10": 0.8359375, "reward/mean_r_ndcg": 0.7431640625, "reward/mean_r_ndcg_cutoff": 5.5, "reward/mean_recall": 0.207275390625, "reward/mean_recall_at_10": 0.04754638671875, "reward/mean_recall_at_100": 0.207275390625, "reward/mean_retrieval_combined_bounded": 0.04241943359375, "reward/mean_retrieval_combined_raw": 0.070709228515625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9892578125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9892578125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 752.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.001953125, "reward/too_short_query_rate": 0.0, "reward_std": 0.28035689145326614, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.2802734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5699549317359924, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.10118421912193298, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.080698728561401, "sampling/raw_vllm_sampling_logp_difference/mean": 0.35553717613220215, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 755, "step_time": 14.073684148781467 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.75390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.8, "completions/min_terminated_length": 0.0, "entropy": 0.48921996653079985, "epoch": 3.0039525691699605, "frac_reward_zero_std": 0.0, "grad_norm": 1.795670747756958, "learning_rate": 2e-05, "loss": 0.0028076171875, "num_tokens": 45445634.0, "reward": 1.016033935546875, "reward/completion_length_max": 39.8, "reward/completion_length_mean": 33.75, "reward/completion_length_min": 29.8, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 37.2, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.01171875, "reward/length_out_of_range_rate": 0.0046875, "reward/low_length_query_rate": 0.0046875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01640625, "reward/mean_colbert": 0.96328125, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.8703125, "reward/mean_r_ndcg": 0.7484375, "reward/mean_r_ndcg_cutoff": 6.1375, "reward/mean_recall": 0.215234375, "reward/mean_recall_at_10": 0.048486328125, "reward/mean_recall_at_100": 0.215234375, "reward/mean_retrieval_combined_bounded": 0.03021240234375, "reward/mean_retrieval_combined_raw": 0.05947611331939697, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 757.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0046875, "reward/too_short_query_rate": 0.0, "reward_std": 0.30881357192993164, "rewards/retrieval_reward/mean": 1.01640625, "rewards/retrieval_reward/std": 0.30859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5903789043426513, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.08154468238353729, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.491157341003419, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3604666471481323, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 760, "step_time": 13.998970577225554 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.75078125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.5002817861735821, "epoch": 3.0237154150197627, "frac_reward_zero_std": 0.0, "grad_norm": 2.297236442565918, "learning_rate": 2e-05, "loss": -8.96453857421875e-05, "num_tokens": 45745251.0, "reward": 1.0022186279296874, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 32.75, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 34.8, "reward/completion_length_p95": 35.4, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00390625, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.8875, "reward/mean_r_ndcg": 0.76484375, "reward/mean_r_ndcg_cutoff": 7.35625, "reward/mean_recall": 0.23515625, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.23515625, "reward/mean_retrieval_combined_bounded": 0.0062160491943359375, "reward/mean_retrieval_combined_raw": 0.02120828628540039, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 761.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2838021695613861, "rewards/retrieval_reward/mean": 1.00390625, "rewards/retrieval_reward/std": 0.283984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 1.8030535221099853, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.05604669600725174, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.445399475097656, "sampling/raw_vllm_sampling_logp_difference/mean": 0.4026742339134216, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 765, "step_time": 13.957817500596866 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.51953125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.5062833152711391, "epoch": 3.0434782608695654, "frac_reward_zero_std": 0.0, "grad_norm": 2.065464496612549, "learning_rate": 2e-05, "loss": -0.0003814697265625, "num_tokens": 46044636.0, "reward": 1.01578369140625, "reward/completion_length_max": 38.8, "reward/completion_length_mean": 32.55, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 35.4, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.9703125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.90234375, "reward/mean_r_ndcg": 0.775, "reward/mean_r_ndcg_cutoff": 6.59375, "reward/mean_recall": 0.2310546875, "reward/mean_recall_at_10": 0.04921875, "reward/mean_recall_at_100": 0.2310546875, "reward/mean_retrieval_combined_bounded": 0.022471237182617187, "reward/mean_retrieval_combined_raw": 0.03238096237182617, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 767.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.28503152132034304, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.28515625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.128297734260559, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.07849522233009339, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.137786293029786, "sampling/raw_vllm_sampling_logp_difference/mean": 0.37503901720046995, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 770, "step_time": 13.934019805223215 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.32421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.4956020303070545, "epoch": 3.0632411067193677, "frac_reward_zero_std": 0.0, "grad_norm": 2.1499969959259033, "learning_rate": 2e-05, "loss": 0.00208740234375, "num_tokens": 46345019.0, "reward": 1.0183624267578124, "reward/completion_length_max": 39.0, "reward/completion_length_mean": 33.3, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 36.8, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.003125, "reward/low_length_query_rate": 0.003125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.9703125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.9046875, "reward/mean_r_ndcg": 0.7796875, "reward/mean_r_ndcg_cutoff": 6.15, "reward/mean_recall": 0.233984375, "reward/mean_recall_at_10": 0.04892578125, "reward/mean_recall_at_100": 0.233984375, "reward/mean_retrieval_combined_bounded": 0.02608642578125, "reward/mean_retrieval_combined_raw": 0.04255943298339844, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 772.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.003125, "reward/too_short_query_rate": 0.0, "reward_std": 0.28735883831977843, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.28828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.539702320098877, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.07045066356658936, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.715742301940917, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3605398952960968, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 775, "step_time": 13.9870178793557 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.53125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.49502176120877267, "epoch": 3.08300395256917, "frac_reward_zero_std": 0.0, "grad_norm": 2.234851121902466, "learning_rate": 2e-05, "loss": 0.0016845703125, "num_tokens": 46644099.0, "reward": 1.0158538818359375, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 32.5, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.85859375, "reward/mean_r_ndcg": 0.7125, "reward/mean_r_ndcg_cutoff": 6.425, "reward/mean_recall": 0.25390625, "reward/mean_recall_at_10": 0.048046875, "reward/mean_recall_at_100": 0.25390625, "reward/mean_retrieval_combined_bounded": 0.0169677734375, "reward/mean_retrieval_combined_raw": 0.004585964232683181, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 776.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2840527892112732, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.283984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.625099539756775, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.08171347305178642, "sampling/raw_vllm_importance_sampling_ratio/min": 1.524692847562452e-15, "sampling/raw_vllm_sampling_logp_difference/max": 12.897525215148926, "sampling/raw_vllm_sampling_logp_difference/mean": 0.39855016469955445, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 780, "step_time": 13.941854024026544 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 56.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.98046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.2, "completions/min_terminated_length": 0.0, "entropy": 0.43782589584589005, "epoch": 3.102766798418972, "frac_reward_zero_std": 0.0, "grad_norm": 2.0963892936706543, "learning_rate": 2e-05, "loss": 0.02265625, "num_tokens": 46943754.0, "reward": 1.044854736328125, "reward/completion_length_max": 56.2, "reward/completion_length_mean": 32.975, "reward/completion_length_min": 27.2, "reward/completion_length_p90": 35.4, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.01171875, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0453125, "reward/mean_colbert": 0.965625, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.84296875, "reward/mean_r_ndcg": 0.74375, "reward/mean_r_ndcg_cutoff": 5.765625, "reward/mean_recall": 0.2240234375, "reward/mean_recall_at_10": 0.047021484375, "reward/mean_recall_at_100": 0.2240234375, "reward/mean_retrieval_combined_bounded": 0.0667236328125, "reward/mean_retrieval_combined_raw": 0.17711849212646485, "reward/output_tag_rate": 0.9953125, "reward/query_valid_rate": 0.98828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 782.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.34540767669677735, "rewards/retrieval_reward/mean": 1.0453125, "rewards/retrieval_reward/std": 0.34609375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.678550148010254, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.08954536393284798, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.68277645111084, "sampling/raw_vllm_sampling_logp_difference/mean": 0.34999805092811587, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 785, "step_time": 14.795532169996296 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.3734375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.5175115317106247, "epoch": 3.122529644268775, "frac_reward_zero_std": 0.0, "grad_norm": 1.987423062324524, "learning_rate": 2e-05, "loss": 0.001312255859375, "num_tokens": 47242632.0, "reward": 0.9902099609375, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 32.35, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.01875, "reward/duplicate_ranked_plain_query_rate": 0.01875, "reward/invalid_output_rate": 0.01875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.990625, "reward/mean_colbert": 0.95625, "reward/mean_mrr": 0.971875, "reward/mean_ndcg_at_10": 0.84453125, "reward/mean_r_ndcg": 0.7, "reward/mean_r_ndcg_cutoff": 5.225, "reward/mean_recall": 0.212109375, "reward/mean_recall_at_10": 0.04677734375, "reward/mean_recall_at_100": 0.212109375, "reward/mean_retrieval_combined_bounded": 0.00888671875, "reward/mean_retrieval_combined_raw": 0.028369717299938202, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 786.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2957346737384796, "rewards/retrieval_reward/mean": 0.990625, "rewards/retrieval_reward/std": 0.29609375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3124234676361084, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.08896129503846169, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.740683937072754, "sampling/raw_vllm_sampling_logp_difference/mean": 0.33607096672058107, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 790, "step_time": 13.865007236541715 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.178125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.5301575213670731, "epoch": 3.142292490118577, "frac_reward_zero_std": 0.0, "grad_norm": 1.533118486404419, "learning_rate": 2e-05, "loss": -0.001470947265625, "num_tokens": 47542732.0, "reward": 1.03740234375, "reward/completion_length_max": 40.2, "reward/completion_length_mean": 33.15, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.6, "reward/duplicate_query_rate": 0.01328125, "reward/duplicate_ranked_plain_query_rate": 0.01328125, "reward/invalid_output_rate": 0.0140625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0390625, "reward/mean_colbert": 0.96015625, "reward/mean_mrr": 0.98359375, "reward/mean_ndcg_at_10": 0.8609375, "reward/mean_r_ndcg": 0.709375, "reward/mean_r_ndcg_cutoff": 5.103125, "reward/mean_recall": 0.207421875, "reward/mean_recall_at_10": 0.047021484375, "reward/mean_recall_at_100": 0.207421875, "reward/mean_retrieval_combined_bounded": 0.0517822265625, "reward/mean_retrieval_combined_raw": 0.141943359375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9859375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9859375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 792.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2808200031518936, "rewards/retrieval_reward/mean": 1.0390625, "rewards/retrieval_reward/std": 0.280859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.450481986999512, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.10098830461502076, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.944753265380859, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3175406396389008, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 795, "step_time": 14.046416204539128 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.76328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.8, "completions/min_terminated_length": 0.0, "entropy": 0.5000106722116471, "epoch": 3.1620553359683794, "frac_reward_zero_std": 0.0, "grad_norm": 1.7804853916168213, "learning_rate": 2e-05, "loss": 0.00201416015625, "num_tokens": 47843389.0, "reward": 1.0074066162109374, "reward/completion_length_max": 40.4, "reward/completion_length_mean": 33.8, "reward/completion_length_min": 29.8, "reward/completion_length_p90": 36.4, "reward/completion_length_p95": 37.8, "reward/duplicate_query_rate": 0.00859375, "reward/duplicate_ranked_plain_query_rate": 0.00859375, "reward/invalid_output_rate": 0.0234375, "reward/length_out_of_range_rate": 0.01484375, "reward/low_length_query_rate": 0.01484375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00625, "reward/mean_colbert": 0.95390625, "reward/mean_mrr": 0.9765625, "reward/mean_ndcg_at_10": 0.878125, "reward/mean_r_ndcg": 0.7171875, "reward/mean_r_ndcg_cutoff": 5.38125, "reward/mean_recall": 0.21953125, "reward/mean_recall_at_10": 0.047216796875, "reward/mean_recall_at_100": 0.21953125, "reward/mean_retrieval_combined_bounded": 0.03826904296875, "reward/mean_retrieval_combined_raw": 0.07597656587604433, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 797.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01484375, "reward/too_short_query_rate": 0.0, "reward_std": 0.3094992101192474, "rewards/retrieval_reward/mean": 1.00625, "rewards/retrieval_reward/std": 0.3095703125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.654199743270874, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.10573760420084, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.182686233520508, "sampling/raw_vllm_sampling_logp_difference/mean": 0.35194127559661864, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 800, "step_time": 14.110355002153664 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.75390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.5247130244970322, "epoch": 3.16600790513834, "frac_reward_zero_std": 0.0, "num_tokens": 47903422.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6628241913634555, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7153008455695532, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8945518453427065, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9929701230228472, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6035316764030804, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6586328863916374, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8466666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3336014493856292, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.36938017811021306, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5228915662650603, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8337349397590361, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6082104487928787, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6206772093163639, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7508505578928114, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9112969483568075, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5730277447007142, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3585735474795065, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17714654182867545, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.312284790755219, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 4.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3448.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.41144440036140806, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4489069054810266, "retrieval_eval/nq/single_turn/recall_at_10": 0.651802977571539, "retrieval_eval/nq/single_turn/recall_at_100": 0.8817188708430007, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6692363840341954, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6981133919432385, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8227777777777778, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9483333333333334, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0195465087890625, "reward/completion_length_max": 40.0, "reward/completion_length_mean": 33.75, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 39.0, "reward/completion_length_p95": 39.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.015625, "reward/malformed_ranked_plain_query_rate": 0.015625, "reward/mean": 1.015625, "reward/mean_colbert": 0.95703125, "reward/mean_mrr": 0.98046875, "reward/mean_ndcg_at_10": 0.859375, "reward/mean_r_ndcg": 0.76171875, "reward/mean_r_ndcg_cutoff": 6.34375, "reward/mean_recall": 0.2470703125, "reward/mean_recall_at_10": 0.049072265625, "reward/mean_recall_at_100": 0.2470703125, "reward/mean_retrieval_combined_bounded": 0.0546875, "reward/mean_retrieval_combined_raw": 0.0927734375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 800.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3863318860530853, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.388671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8300187587738037, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15896545350551605, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.52833080291748, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2755794823169708, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 801, "step_time": 14.130341169075109 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.3486328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.5, "completions/min_terminated_length": 0.0, "entropy": 0.4684506254270673, "epoch": 3.1818181818181817, "frac_reward_zero_std": 0.0, "grad_norm": 2.359468460083008, "learning_rate": 2e-05, "loss": -0.000799560546875, "num_tokens": 48143587.0, "reward": 1.0008125305175781, "reward/completion_length_max": 39.25, "reward/completion_length_mean": 33.375, "reward/completion_length_min": 29.5, "reward/completion_length_p90": 35.5, "reward/completion_length_p95": 36.5, "reward/duplicate_query_rate": 0.0244140625, "reward/duplicate_ranked_plain_query_rate": 0.0244140625, "reward/invalid_output_rate": 0.0244140625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.001953125, "reward/mean_colbert": 0.9501953125, "reward/mean_mrr": 0.9755859375, "reward/mean_ndcg_at_10": 0.837890625, "reward/mean_r_ndcg": 0.66796875, "reward/mean_r_ndcg_cutoff": 6.4609375, "reward/mean_recall": 0.210693359375, "reward/mean_recall_at_10": 0.047607421875, "reward/mean_recall_at_100": 0.210693359375, "reward/mean_retrieval_combined_bounded": 0.025115966796875, "reward/mean_retrieval_combined_raw": 0.04150392464362085, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9755859375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9755859375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 802.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.33056434243917465, "rewards/retrieval_reward/mean": 1.001953125, "rewards/retrieval_reward/std": 0.3310546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.61284077167511, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12376260198652744, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.083384990692139, "sampling/raw_vllm_sampling_logp_difference/mean": 0.29637330770492554, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 805, "step_time": 13.993826312260353 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.62109375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.8, "completions/min_terminated_length": 0.0, "entropy": 0.5120889745652676, "epoch": 3.2015810276679844, "frac_reward_zero_std": 0.0, "grad_norm": 1.9756561517715454, "learning_rate": 2e-05, "loss": 0.0019775390625, "num_tokens": 48444254.0, "reward": 1.019122314453125, "reward/completion_length_max": 40.4, "reward/completion_length_mean": 33.65, "reward/completion_length_min": 29.8, "reward/completion_length_p90": 36.4, "reward/completion_length_p95": 37.6, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.00625, "reward/low_length_query_rate": 0.00625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.96640625, "reward/mean_mrr": 0.990625, "reward/mean_ndcg_at_10": 0.90078125, "reward/mean_r_ndcg": 0.7625, "reward/mean_r_ndcg_cutoff": 6.1625, "reward/mean_recall": 0.218359375, "reward/mean_recall_at_10": 0.048388671875, "reward/mean_recall_at_100": 0.218359375, "reward/mean_retrieval_combined_bounded": 0.031744384765625, "reward/mean_retrieval_combined_raw": 0.049118983745574954, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 807.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00625, "reward/too_short_query_rate": 0.0, "reward_std": 0.29632446765899656, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.296484375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4545600414276123, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.10212972015142441, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.537665367126465, "sampling/raw_vllm_sampling_logp_difference/mean": 0.32098196148872377, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 810, "step_time": 14.105764938390347 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.99921875, "completions/max_length": 38.6, "completions/max_terminated_length": 5.4, "completions/mean_length": 33.6890625, "completions/mean_terminated_length": 5.4, "completions/min_length": 29.0, "completions/min_terminated_length": 5.4, "entropy": 0.4917343966662884, "epoch": 3.2213438735177866, "frac_reward_zero_std": 0.0, "grad_norm": 1.9573527574539185, "learning_rate": 2e-05, "loss": -0.001446533203125, "num_tokens": 48744976.0, "reward": 1.025933837890625, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 33.7, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.00859375, "reward/duplicate_ranked_plain_query_rate": 0.00859375, "reward/invalid_output_rate": 0.01015625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.98984375, "reward/mean_ndcg_at_10": 0.8859375, "reward/mean_r_ndcg": 0.721875, "reward/mean_r_ndcg_cutoff": 7.1125, "reward/mean_recall": 0.2390625, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.2390625, "reward/mean_retrieval_combined_bounded": 0.03809814453125, "reward/mean_retrieval_combined_raw": 0.06430678591132163, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.98984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 812.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2804442703723907, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.28125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.472297525405884, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.108356873691082, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.708735847473145, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3090890824794769, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 815, "step_time": 13.912556803470943 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.6453125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.8, "completions/min_terminated_length": 0.0, "entropy": 0.49398961663246155, "epoch": 3.241106719367589, "frac_reward_zero_std": 0.0, "grad_norm": 1.8494559526443481, "learning_rate": 2e-05, "loss": -0.00033111572265625, "num_tokens": 49044554.0, "reward": 1.0209228515625, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 32.65, "reward/completion_length_min": 28.8, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.009375, "reward/duplicate_ranked_plain_query_rate": 0.009375, "reward/invalid_output_rate": 0.0109375, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.9671875, "reward/mean_mrr": 0.9890625, "reward/mean_ndcg_at_10": 0.9046875, "reward/mean_r_ndcg": 0.78203125, "reward/mean_r_ndcg_cutoff": 6.05625, "reward/mean_recall": 0.230859375, "reward/mean_recall_at_10": 0.048876953125, "reward/mean_recall_at_100": 0.230859375, "reward/mean_retrieval_combined_bounded": 0.03258934020996094, "reward/mean_retrieval_combined_raw": 0.04946333821862936, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 816.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.28089924454689025, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.28125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4505979776382447, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1063517987728119, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.364679718017578, "sampling/raw_vllm_sampling_logp_difference/mean": 0.32099775075912473, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 820, "step_time": 13.975040961639024 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.61015625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.4679192379117012, "epoch": 3.260869565217391, "frac_reward_zero_std": 0.0, "grad_norm": 1.8945854902267456, "learning_rate": 2e-05, "loss": -0.0001312255859375, "num_tokens": 49342231.0, "reward": 1.0154052734375, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 31.6, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 35.4, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01640625, "reward/mean_colbert": 0.9703125, "reward/mean_mrr": 0.99453125, "reward/mean_ndcg_at_10": 0.89453125, "reward/mean_r_ndcg": 0.76328125, "reward/mean_r_ndcg_cutoff": 6.49375, "reward/mean_recall": 0.2341796875, "reward/mean_recall_at_10": 0.048779296875, "reward/mean_recall_at_100": 0.2341796875, "reward/mean_retrieval_combined_bounded": 0.020111083984375, "reward/mean_retrieval_combined_raw": 0.02932043671607971, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 822.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2844890296459198, "rewards/retrieval_reward/mean": 1.01640625, "rewards/retrieval_reward/std": 0.284375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4531058788299562, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13565004765987396, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.680642318725585, "sampling/raw_vllm_sampling_logp_difference/mean": 0.30057480931282043, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 825, "step_time": 13.865648977004458 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.521875, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.8, "completions/min_terminated_length": 0.0, "entropy": 0.4385067895054817, "epoch": 3.280632411067194, "frac_reward_zero_std": 0.0, "grad_norm": 1.8060725927352905, "learning_rate": 2e-05, "loss": -0.0016357421875, "num_tokens": 49638963.0, "reward": 1.0175750732421875, "reward/completion_length_max": 35.8, "reward/completion_length_mean": 30.525, "reward/completion_length_min": 26.8, "reward/completion_length_p90": 32.8, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.88046875, "reward/mean_r_ndcg": 0.6765625, "reward/mean_r_ndcg_cutoff": 5.84375, "reward/mean_recall": 0.22421875, "reward/mean_recall_at_10": 0.04873046875, "reward/mean_recall_at_100": 0.22421875, "reward/mean_retrieval_combined_bounded": 0.0190704345703125, "reward/mean_retrieval_combined_raw": 0.012088007107377053, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 826.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.29195359349250793, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.292578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5491084098815917, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.137536358833313, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.858114242553711, "sampling/raw_vllm_sampling_logp_difference/mean": 0.30227890610694885, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 830, "step_time": 13.746369309804868 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.36953125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.44257529452443123, "epoch": 3.300395256916996, "frac_reward_zero_std": 0.0, "grad_norm": 2.099525213241577, "learning_rate": 2e-05, "loss": -0.00177001953125, "num_tokens": 49934828.0, "reward": 1.02760009765625, "reward/completion_length_max": 37.6, "reward/completion_length_mean": 30.35, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 33.4, "reward/completion_length_p95": 34.2, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.00546875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.99453125, "reward/mean_ndcg_at_10": 0.88671875, "reward/mean_r_ndcg": 0.74609375, "reward/mean_r_ndcg_cutoff": 5.98125, "reward/mean_recall": 0.2259765625, "reward/mean_recall_at_10": 0.048193359375, "reward/mean_recall_at_100": 0.2259765625, "reward/mean_retrieval_combined_bounded": 0.0329833984375, "reward/mean_retrieval_combined_raw": 0.030200260132551192, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99453125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99453125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 832.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28697946667671204, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.287109375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5898849248886107, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1523812249302864, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.318841171264648, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2930573642253876, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 835, "step_time": 13.848723235039506 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.74453125, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.4, "completions/min_terminated_length": 0.0, "entropy": 0.45365139096975327, "epoch": 3.3201581027667983, "frac_reward_zero_std": 0.0, "grad_norm": 1.8616384267807007, "learning_rate": 2e-05, "loss": 0.000750732421875, "num_tokens": 50230565.0, "reward": 1.016009521484375, "reward/completion_length_max": 35.6, "reward/completion_length_mean": 29.725, "reward/completion_length_min": 26.4, "reward/completion_length_p90": 32.2, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.90703125, "reward/mean_r_ndcg": 0.73984375, "reward/mean_r_ndcg_cutoff": 6.56875, "reward/mean_recall": 0.2169921875, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.2169921875, "reward/mean_retrieval_combined_bounded": 0.01990966796875, "reward/mean_retrieval_combined_raw": 0.018612544797360898, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 837.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28699966669082644, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.28671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.552301597595215, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13515894263982772, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.522926330566406, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27884837687015535, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 840, "step_time": 13.791065481200349 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.13125, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.4, "completions/min_terminated_length": 0.0, "entropy": 0.44495713263750075, "epoch": 3.3399209486166006, "frac_reward_zero_std": 0.0, "grad_norm": 1.6457794904708862, "learning_rate": 2e-05, "loss": -0.0004486083984375, "num_tokens": 50526637.0, "reward": 1.034234619140625, "reward/completion_length_max": 35.6, "reward/completion_length_mean": 30.125, "reward/completion_length_min": 26.4, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.034375, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.8734375, "reward/mean_r_ndcg": 0.7640625, "reward/mean_r_ndcg_cutoff": 6.0875, "reward/mean_recall": 0.2478515625, "reward/mean_recall_at_10": 0.048486328125, "reward/mean_recall_at_100": 0.2478515625, "reward/mean_retrieval_combined_bounded": 0.04129638671875, "reward/mean_retrieval_combined_raw": 0.07109360694885254, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 841.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2824856460094452, "rewards/retrieval_reward/mean": 1.034375, "rewards/retrieval_reward/std": 0.2830078125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8275925636291506, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16688761711120606, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.5785888671875, "sampling/raw_vllm_sampling_logp_difference/mean": 0.29211120009422303, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 845, "step_time": 13.734851900010835 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.9, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.6, "completions/min_terminated_length": 0.0, "entropy": 0.44453613460063934, "epoch": 3.3596837944664033, "frac_reward_zero_std": 0.0, "grad_norm": 1.8818552494049072, "learning_rate": 2e-05, "loss": -0.000811767578125, "num_tokens": 50820973.0, "reward": 1.0170989990234376, "reward/completion_length_max": 32.8, "reward/completion_length_mean": 28.9, "reward/completion_length_min": 24.6, "reward/completion_length_p90": 30.8, "reward/completion_length_p95": 31.4, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.91015625, "reward/mean_r_ndcg": 0.8, "reward/mean_r_ndcg_cutoff": 6.18125, "reward/mean_recall": 0.23984375, "reward/mean_recall_at_10": 0.049072265625, "reward/mean_recall_at_100": 0.23984375, "reward/mean_retrieval_combined_bounded": 0.02218017578125, "reward/mean_retrieval_combined_raw": 0.02234072987921536, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 847.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00234375, "reward_std": 0.2954627752304077, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.29609375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6651614189147947, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13402241319417954, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.14010829925537, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3175202548503876, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 850, "step_time": 13.631599992583507 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.5859375, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.0, "completions/min_terminated_length": 0.0, "entropy": 0.4304171986877918, "epoch": 3.3636363636363638, "frac_reward_zero_std": 0.0, "num_tokens": 50879875.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6578872703813079, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.710146127203362, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8910369068541301, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9956063268892794, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6093265468340141, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6610346255176109, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8466666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 1.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3425012570250871, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3803821185245344, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5421686746987951, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8313253012048193, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6225178146586637, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6233461878501337, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7381886013047985, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9027944180232913, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5733925695863726, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.359318964486813, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17815036207235566, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3115015172747028, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 0.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3452.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.3992635980940544, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4362594339239668, "retrieval_eval/nq/single_turn/recall_at_10": 0.6371427191966009, "retrieval_eval/nq/single_turn/recall_at_100": 0.875, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6833348320866741, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7099890942353265, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8380555555555556, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9493333333333334, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.027923583984375, "reward/completion_length_max": 33.0, "reward/completion_length_mean": 28.625, "reward/completion_length_min": 24.0, "reward/completion_length_p90": 31.0, "reward/completion_length_p95": 31.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.86328125, "reward/mean_r_ndcg": 0.76953125, "reward/mean_r_ndcg_cutoff": 6.0625, "reward/mean_recall": 0.2255859375, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.2255859375, "reward/mean_retrieval_combined_bounded": 0.03759765625, "reward/mean_retrieval_combined_raw": 0.06396484375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 848.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00390625, "reward_std": 0.24025388062000275, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.240234375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6365420818328857, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21706417202949524, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.020401954650879, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2520284056663513, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 851, "step_time": 13.54094029602129 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.2822265625, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.75, "completions/min_terminated_length": 0.0, "entropy": 0.437495824880898, "epoch": 3.3794466403162056, "frac_reward_zero_std": 0.0, "grad_norm": 1.9823377132415771, "learning_rate": 2e-05, "loss": 0.000445556640625, "num_tokens": 51114820.0, "reward": 1.0351371765136719, "reward/completion_length_max": 32.25, "reward/completion_length_mean": 28.28125, "reward/completion_length_min": 24.75, "reward/completion_length_p90": 30.5, "reward/completion_length_p95": 30.75, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.001953125, "reward/length_out_of_range_rate": 0.001953125, "reward/low_length_query_rate": 0.001953125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.033203125, "reward/mean_colbert": 0.9755859375, "reward/mean_mrr": 0.998046875, "reward/mean_ndcg_at_10": 0.8994140625, "reward/mean_r_ndcg": 0.7890625, "reward/mean_r_ndcg_cutoff": 6.234375, "reward/mean_recall": 0.20947265625, "reward/mean_recall_at_10": 0.04913330078125, "reward/mean_recall_at_100": 0.20947265625, "reward/mean_retrieval_combined_bounded": 0.0379791259765625, "reward/mean_retrieval_combined_raw": 0.06213286519050598, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.998046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.998046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 853.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.001953125, "reward_std": 0.2882639989256859, "rewards/retrieval_reward/mean": 1.033203125, "rewards/retrieval_reward/std": 0.28857421875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8035168647766113, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16267018765211105, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.736101627349854, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3042062297463417, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 855, "step_time": 13.60823693999555 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 27.63984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.2, "completions/min_terminated_length": 0.0, "entropy": 0.40633328780531885, "epoch": 3.399209486166008, "frac_reward_zero_std": 0.0, "grad_norm": 2.1996617317199707, "learning_rate": 2e-05, "loss": -0.0057861328125, "num_tokens": 51407479.0, "reward": 0.993756103515625, "reward/completion_length_max": 32.6, "reward/completion_length_mean": 27.625, "reward/completion_length_min": 24.2, "reward/completion_length_p90": 29.8, "reward/completion_length_p95": 30.6, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.028125, "reward/length_out_of_range_rate": 0.028125, "reward/low_length_query_rate": 0.028125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.9921875, "reward/mean_colbert": 0.94765625, "reward/mean_mrr": 0.971875, "reward/mean_ndcg_at_10": 0.86171875, "reward/mean_r_ndcg": 0.77421875, "reward/mean_r_ndcg_cutoff": 5.86875, "reward/mean_recall": 0.2115234375, "reward/mean_recall_at_10": 0.047509765625, "reward/mean_recall_at_100": 0.2115234375, "reward/mean_retrieval_combined_bounded": 0.0357421875, "reward/mean_retrieval_combined_raw": 0.0570320725440979, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.971875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.971875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 856.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.028125, "reward_std": 0.34979432821273804, "rewards/retrieval_reward/mean": 0.9921875, "rewards/retrieval_reward/std": 0.350390625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7161810874938963, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1828143447637558, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.940532302856445, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2905315935611725, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 860, "step_time": 13.520479200803674 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 31.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 27.14921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.0, "completions/min_terminated_length": 0.0, "entropy": 0.42299827486276625, "epoch": 3.4189723320158105, "frac_reward_zero_std": 0.0, "grad_norm": 1.754446268081665, "learning_rate": 2e-05, "loss": -0.0052734375, "num_tokens": 51699446.0, "reward": 0.99591064453125, "reward/completion_length_max": 31.8, "reward/completion_length_mean": 27.15, "reward/completion_length_min": 24.0, "reward/completion_length_p90": 29.0, "reward/completion_length_p95": 29.8, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.03671875, "reward/length_out_of_range_rate": 0.0359375, "reward/low_length_query_rate": 0.0359375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99453125, "reward/mean_colbert": 0.9390625, "reward/mean_mrr": 0.9625, "reward/mean_ndcg_at_10": 0.86015625, "reward/mean_r_ndcg": 0.74140625, "reward/mean_r_ndcg_cutoff": 5.61875, "reward/mean_recall": 0.1875, "reward/mean_recall_at_10": 0.047119140625, "reward/mean_recall_at_100": 0.1875, "reward/mean_retrieval_combined_bounded": 0.0505126953125, "reward/mean_retrieval_combined_raw": 0.0980712890625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.96328125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96328125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 862.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0359375, "reward_std": 0.3630812048912048, "rewards/retrieval_reward/mean": 0.99453125, "rewards/retrieval_reward/std": 0.3640625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.554865574836731, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.17194412052631378, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.588747024536133, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2844027131795883, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 865, "step_time": 13.454249554418492 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.203125, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.6, "completions/min_terminated_length": 0.0, "entropy": 0.4533442914485931, "epoch": 3.438735177865613, "frac_reward_zero_std": 0.0, "grad_norm": 2.8928983211517334, "learning_rate": 2e-05, "loss": -0.001239013671875, "num_tokens": 51992602.0, "reward": 1.01865234375, "reward/completion_length_max": 33.0, "reward/completion_length_mean": 28.2, "reward/completion_length_min": 24.6, "reward/completion_length_p90": 30.8, "reward/completion_length_p95": 31.2, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.00546875, "reward/low_length_query_rate": 0.00546875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.8796875, "reward/mean_r_ndcg": 0.7546875, "reward/mean_r_ndcg_cutoff": 6.29375, "reward/mean_recall": 0.219140625, "reward/mean_recall_at_10": 0.049072265625, "reward/mean_recall_at_100": 0.219140625, "reward/mean_retrieval_combined_bounded": 0.0282958984375, "reward/mean_retrieval_combined_raw": 0.03554758671671152, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 866.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00546875, "reward_std": 0.2943451404571533, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.294921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7310011863708494, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1634287193417549, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.627386474609375, "sampling/raw_vllm_sampling_logp_difference/mean": 0.29845130443573, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 870, "step_time": 13.577264154225123 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.67890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.8, "completions/min_terminated_length": 0.0, "entropy": 0.45236233472824094, "epoch": 3.458498023715415, "frac_reward_zero_std": 0.0, "grad_norm": 2.5763447284698486, "learning_rate": 2e-05, "loss": -0.00023956298828125, "num_tokens": 52287007.0, "reward": 1.0159393310546876, "reward/completion_length_max": 33.6, "reward/completion_length_mean": 28.675, "reward/completion_length_min": 25.8, "reward/completion_length_p90": 30.4, "reward/completion_length_p95": 31.0, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.84921875, "reward/mean_r_ndcg": 0.6734375, "reward/mean_r_ndcg_cutoff": 6.00625, "reward/mean_recall": 0.2330078125, "reward/mean_recall_at_10": 0.049072265625, "reward/mean_recall_at_100": 0.2330078125, "reward/mean_retrieval_combined_bounded": 0.02054443359375, "reward/mean_retrieval_combined_raw": 0.02050764635205269, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 872.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28057320713996886, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.28125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7531754970550537, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13681662529706956, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.029720497131347, "sampling/raw_vllm_sampling_logp_difference/mean": 0.301452112197876, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 875, "step_time": 13.619493703835179 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 27.99609375, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.0, "completions/min_terminated_length": 0.0, "entropy": 0.3927656836807728, "epoch": 3.4782608695652173, "frac_reward_zero_std": 0.0, "grad_norm": 1.9723145961761475, "learning_rate": 2e-05, "loss": -0.00260009765625, "num_tokens": 52580762.0, "reward": 1.0223175048828126, "reward/completion_length_max": 32.4, "reward/completion_length_mean": 28.0, "reward/completion_length_min": 24.0, "reward/completion_length_p90": 30.2, "reward/completion_length_p95": 30.8, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0109375, "reward/length_out_of_range_rate": 0.009375, "reward/low_length_query_rate": 0.009375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.9625, "reward/mean_mrr": 0.9875, "reward/mean_ndcg_at_10": 0.884375, "reward/mean_r_ndcg": 0.74453125, "reward/mean_r_ndcg_cutoff": 6.05625, "reward/mean_recall": 0.2267578125, "reward/mean_recall_at_10": 0.0486328125, "reward/mean_recall_at_100": 0.2267578125, "reward/mean_retrieval_combined_bounded": 0.037945556640625, "reward/mean_retrieval_combined_raw": 0.05150470733642578, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 877.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.009375, "reward_std": 0.30511019825935365, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.30546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.758046817779541, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1802251845598221, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 13.680620765686035, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2905225545167923, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 880, "step_time": 13.574374783749226 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 26.88359375, "completions/mean_terminated_length": 0.0, "completions/min_length": 23.6, "completions/min_terminated_length": 0.0, "entropy": 0.33900736421346667, "epoch": 3.4980237154150196, "frac_reward_zero_std": 0.025, "grad_norm": 3.5296013355255127, "learning_rate": 2e-05, "loss": -0.0058837890625, "num_tokens": 52872901.0, "reward": 0.8882568359375, "reward/completion_length_max": 33.6, "reward/completion_length_mean": 26.875, "reward/completion_length_min": 23.6, "reward/completion_length_p90": 30.0, "reward/completion_length_p95": 31.4, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.11875, "reward/length_out_of_range_rate": 0.115625, "reward/low_length_query_rate": 0.115625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.8875, "reward/mean_colbert": 0.859375, "reward/mean_mrr": 0.88125, "reward/mean_ndcg_at_10": 0.78984375, "reward/mean_r_ndcg": 0.676953125, "reward/mean_r_ndcg_cutoff": 5.4125, "reward/mean_recall": 0.1958984375, "reward/mean_recall_at_10": 0.04296875, "reward/mean_recall_at_100": 0.1958984375, "reward/mean_retrieval_combined_bounded": 0.064837646484375, "reward/mean_retrieval_combined_raw": 0.11103515625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.88125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.88125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 881.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.115625, "reward_std": 0.5049563765525817, "rewards/retrieval_reward/mean": 0.88828125, "rewards/retrieval_reward/std": 0.505859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8910217761993406, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.189754918217659, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.272577285766602, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27942569851875304, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 885, "step_time": 13.419408076617401 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 26.88671875, "completions/mean_terminated_length": 0.0, "completions/min_length": 23.6, "completions/min_terminated_length": 0.0, "entropy": 0.311840382963419, "epoch": 3.5177865612648223, "frac_reward_zero_std": 0.0, "grad_norm": 2.5532748699188232, "learning_rate": 2e-05, "loss": -0.0057861328125, "num_tokens": 53164756.0, "reward": 0.9480560302734375, "reward/completion_length_max": 32.4, "reward/completion_length_mean": 26.9, "reward/completion_length_min": 23.6, "reward/completion_length_p90": 29.2, "reward/completion_length_p95": 30.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.0765625, "reward/length_out_of_range_rate": 0.07265625, "reward/low_length_query_rate": 0.07265625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.94765625, "reward/mean_colbert": 0.90234375, "reward/mean_mrr": 0.9234375, "reward/mean_ndcg_at_10": 0.8578125, "reward/mean_r_ndcg": 0.784375, "reward/mean_r_ndcg_cutoff": 6.3, "reward/mean_recall": 0.217578125, "reward/mean_recall_at_10": 0.04560546875, "reward/mean_recall_at_100": 0.217578125, "reward/mean_retrieval_combined_bounded": 0.0611572265625, "reward/mean_retrieval_combined_raw": 0.14327774047851563, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9234375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9234375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 887.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.07265625, "reward_std": 0.4614567160606384, "rewards/retrieval_reward/mean": 0.946875, "rewards/retrieval_reward/std": 0.462109375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.841834306716919, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21148669719696045, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.370465087890626, "sampling/raw_vllm_sampling_logp_difference/mean": 0.273964923620224, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 890, "step_time": 13.406685891747475 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 27.84375, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.8, "completions/min_terminated_length": 0.0, "entropy": 0.4094063565135002, "epoch": 3.5375494071146245, "frac_reward_zero_std": 0.0, "grad_norm": 2.193077802658081, "learning_rate": 2e-05, "loss": -0.00164794921875, "num_tokens": 53457964.0, "reward": 1.015643310546875, "reward/completion_length_max": 33.0, "reward/completion_length_mean": 27.85, "reward/completion_length_min": 24.8, "reward/completion_length_p90": 30.0, "reward/completion_length_p95": 30.4, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.00703125, "reward/low_length_query_rate": 0.00703125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.9640625, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.89453125, "reward/mean_r_ndcg": 0.75625, "reward/mean_r_ndcg_cutoff": 5.73125, "reward/mean_recall": 0.1982421875, "reward/mean_recall_at_10": 0.047802734375, "reward/mean_recall_at_100": 0.1982421875, "reward/mean_retrieval_combined_bounded": 0.0284912109375, "reward/mean_retrieval_combined_raw": 0.04890190064907074, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 892.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00703125, "reward_std": 0.2859056144952774, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.286328125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7174516677856446, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21171716451644898, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.323715019226075, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26314362287521365, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 895, "step_time": 13.554085786186624 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.10703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.6, "completions/min_terminated_length": 0.0, "entropy": 0.41224599480628965, "epoch": 3.5573122529644268, "frac_reward_zero_std": 0.0, "grad_norm": 1.8514697551727295, "learning_rate": 2e-05, "loss": -0.00228271484375, "num_tokens": 53752789.0, "reward": 1.0202484130859375, "reward/completion_length_max": 33.6, "reward/completion_length_mean": 29.1, "reward/completion_length_min": 25.6, "reward/completion_length_p90": 31.2, "reward/completion_length_p95": 32.0, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.99765625, "reward/mean_ndcg_at_10": 0.88046875, "reward/mean_r_ndcg": 0.75625, "reward/mean_r_ndcg_cutoff": 6.38125, "reward/mean_recall": 0.2453125, "reward/mean_recall_at_10": 0.04912109375, "reward/mean_recall_at_100": 0.2453125, "reward/mean_retrieval_combined_bounded": 0.0225341796875, "reward/mean_retrieval_combined_raw": 0.018816915154457093, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 896.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.26702251136302946, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.2671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8530741214752195, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1923963099718094, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.815909385681152, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2684617072343826, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 900, "step_time": 13.608225416135975 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.6328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.47474562004208565, "epoch": 3.561264822134387, "frac_reward_zero_std": 0.0, "num_tokens": 53812183.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6612156151229333, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.712605380541143, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8892794376098418, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9947275922671354, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5772125754488718, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6378445387869471, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8466666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 1.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.33754117712250103, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3711900492724438, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5228915662650603, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8361445783132531, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6145499063217222, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6290704705471611, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7560903298579355, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9113640174379611, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5739741694357077, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3600075667140826, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17770685242963466, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3166996344167399, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 1.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3451.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.41543564083305196, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.45081161717922347, "retrieval_eval/nq/single_turn/recall_at_10": 0.6471312662996233, "retrieval_eval/nq/single_turn/recall_at_100": 0.886482179078528, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6658299925076756, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7001960883594024, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8437777777777777, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9453333333333334, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.01971435546875, "reward/completion_length_max": 34.0, "reward/completion_length_mean": 30.625, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.9765625, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.92578125, "reward/mean_r_ndcg": 0.76171875, "reward/mean_r_ndcg_cutoff": 5.125, "reward/mean_recall": 0.19140625, "reward/mean_recall_at_10": 0.049560546875, "reward/mean_recall_at_100": 0.19140625, "reward/mean_retrieval_combined_bounded": 0.0194091796875, "reward/mean_retrieval_combined_raw": 7.104873657226562e-05, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 900.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27948060631752014, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.279296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.9378652572631836, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2041819840669632, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 5.114204406738281, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2399267852306366, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 901, "step_time": 13.615263525978662 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.75, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.1162109375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.5, "completions/min_terminated_length": 0.0, "entropy": 0.460046648979187, "epoch": 3.5770750988142295, "frac_reward_zero_std": 0.0, "grad_norm": 1.9535207748413086, "learning_rate": 2e-05, "loss": 0.0002960205078125, "num_tokens": 54049998.0, "reward": 1.0149078369140625, "reward/completion_length_max": 36.75, "reward/completion_length_mean": 31.125, "reward/completion_length_min": 27.5, "reward/completion_length_p90": 33.25, "reward/completion_length_p95": 34.25, "reward/duplicate_query_rate": 0.009765625, "reward/duplicate_ranked_plain_query_rate": 0.009765625, "reward/invalid_output_rate": 0.009765625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.9677734375, "reward/mean_mrr": 0.990234375, "reward/mean_ndcg_at_10": 0.90625, "reward/mean_r_ndcg": 0.794921875, "reward/mean_r_ndcg_cutoff": 6.4609375, "reward/mean_recall": 0.240478515625, "reward/mean_recall_at_10": 0.04852294921875, "reward/mean_recall_at_100": 0.240478515625, "reward/mean_retrieval_combined_bounded": 0.024541854858398438, "reward/mean_retrieval_combined_raw": 0.035178907215595245, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990234375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990234375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 903.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28930098563432693, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.2900390625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.506527602672577, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16570195741951466, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.2805702686309814, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25615841895341873, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 905, "step_time": 13.85055024905887 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.95703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.8, "completions/min_terminated_length": 0.0, "entropy": 0.4683261826634407, "epoch": 3.5968379446640317, "frac_reward_zero_std": 0.0, "grad_norm": 1.8515777587890625, "learning_rate": 2e-05, "loss": 2.1839141845703124e-05, "num_tokens": 54349687.0, "reward": 1.0248809814453126, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 33.0, "reward/completion_length_min": 28.8, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 36.6, "reward/duplicate_query_rate": 0.01953125, "reward/duplicate_ranked_plain_query_rate": 0.01953125, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.02578125, "reward/mean_colbert": 0.95625, "reward/mean_mrr": 0.98046875, "reward/mean_ndcg_at_10": 0.89453125, "reward/mean_r_ndcg": 0.78046875, "reward/mean_r_ndcg_cutoff": 6.43125, "reward/mean_recall": 0.2296875, "reward/mean_recall_at_10": 0.047314453125, "reward/mean_recall_at_100": 0.2296875, "reward/mean_retrieval_combined_bounded": 0.044384765625, "reward/mean_retrieval_combined_raw": 0.07880859375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 906.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3013024151325226, "rewards/retrieval_reward/mean": 1.02578125, "rewards/retrieval_reward/std": 0.301953125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7369804859161375, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.176497583091259, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.795939445495605, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27092020511627196, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 910, "step_time": 13.954658139403909 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.90625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.4581772409379482, "epoch": 3.616600790513834, "frac_reward_zero_std": 0.0, "grad_norm": 1.6497937440872192, "learning_rate": 2e-05, "loss": 0.00184326171875, "num_tokens": 54649151.0, "reward": 1.002142333984375, "reward/completion_length_max": 39.8, "reward/completion_length_mean": 32.875, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.6, "reward/duplicate_query_rate": 0.01328125, "reward/duplicate_ranked_plain_query_rate": 0.01328125, "reward/invalid_output_rate": 0.0265625, "reward/length_out_of_range_rate": 0.01328125, "reward/low_length_query_rate": 0.01328125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00078125, "reward/mean_colbert": 0.95, "reward/mean_mrr": 0.96875, "reward/mean_ndcg_at_10": 0.8671875, "reward/mean_r_ndcg": 0.7859375, "reward/mean_r_ndcg_cutoff": 6.00625, "reward/mean_recall": 0.23359375, "reward/mean_recall_at_10": 0.047216796875, "reward/mean_recall_at_100": 0.23359375, "reward/mean_retrieval_combined_bounded": 0.03531494140625, "reward/mean_retrieval_combined_raw": 0.06977906227111816, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9734375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9734375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 912.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01328125, "reward/too_short_query_rate": 0.0, "reward_std": 0.33331275582313535, "rewards/retrieval_reward/mean": 1.00078125, "rewards/retrieval_reward/std": 0.334375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8606019973754884, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.17172639667987824, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.801611328125, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27420918345451356, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 915, "step_time": 13.931277707440312 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.30390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.45443679913878443, "epoch": 3.6363636363636362, "frac_reward_zero_std": 0.0, "grad_norm": 1.7045197486877441, "learning_rate": 2e-05, "loss": -0.0020751953125, "num_tokens": 54947972.0, "reward": 1.0157989501953124, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 32.375, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.8828125, "reward/mean_r_ndcg": 0.76171875, "reward/mean_r_ndcg_cutoff": 6.79375, "reward/mean_recall": 0.2431640625, "reward/mean_recall_at_10": 0.047998046875, "reward/mean_recall_at_100": 0.2431640625, "reward/mean_retrieval_combined_bounded": 0.022088623046875, "reward/mean_retrieval_combined_raw": 0.032769284397363665, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 917.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28381653428077697, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.284375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5036558151245116, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.14892484843730927, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.24254150390625, "sampling/raw_vllm_sampling_logp_difference/mean": 0.29713705778121946, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 920, "step_time": 13.877795299631543 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.51328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.8, "completions/min_terminated_length": 0.0, "entropy": 0.4840299002826214, "epoch": 3.6561264822134385, "frac_reward_zero_std": 0.0, "grad_norm": 1.8140605688095093, "learning_rate": 2e-05, "loss": -0.00361328125, "num_tokens": 55247189.0, "reward": 1.0225189208984375, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 32.5, "reward/completion_length_min": 27.8, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.99140625, "reward/mean_ndcg_at_10": 0.87265625, "reward/mean_r_ndcg": 0.7421875, "reward/mean_r_ndcg_cutoff": 6.375, "reward/mean_recall": 0.2166015625, "reward/mean_recall_at_10": 0.0466796875, "reward/mean_recall_at_100": 0.2166015625, "reward/mean_retrieval_combined_bounded": 0.0304443359375, "reward/mean_retrieval_combined_raw": 0.03606511354446411, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 921.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2893382728099823, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.290234375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.333842945098877, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.104344642162323, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.625706672668457, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3771235287189484, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 925, "step_time": 13.952162153599783 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.4390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.47802290320396423, "epoch": 3.675889328063241, "frac_reward_zero_std": 0.0, "grad_norm": 1.9389153718948364, "learning_rate": 2e-05, "loss": 0.003515625, "num_tokens": 55547463.0, "reward": 1.01995849609375, "reward/completion_length_max": 41.2, "reward/completion_length_mean": 33.4, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 37.6, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.0234375, "reward/length_out_of_range_rate": 0.01328125, "reward/low_length_query_rate": 0.01328125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.9515625, "reward/mean_mrr": 0.9765625, "reward/mean_ndcg_at_10": 0.88984375, "reward/mean_r_ndcg": 0.7640625, "reward/mean_r_ndcg_cutoff": 6.25, "reward/mean_recall": 0.2041015625, "reward/mean_recall_at_10": 0.047412109375, "reward/mean_recall_at_100": 0.2041015625, "reward/mean_retrieval_combined_bounded": 0.04998779296875, "reward/mean_retrieval_combined_raw": 0.08994206190109252, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 927.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01328125, "reward/too_short_query_rate": 0.0, "reward_std": 0.32737934589385986, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.328125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5453402519226076, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13303584605455399, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.683478927612304, "sampling/raw_vllm_sampling_logp_difference/mean": 0.283709916472435, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 930, "step_time": 14.018314915429801 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.54765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.5144968122243881, "epoch": 3.6956521739130435, "frac_reward_zero_std": 0.0, "grad_norm": 1.9052436351776123, "learning_rate": 2e-05, "loss": -0.0004425048828125, "num_tokens": 55846436.0, "reward": 1.01434326171875, "reward/completion_length_max": 37.2, "reward/completion_length_mean": 32.525, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 34.8, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.0125, "reward/duplicate_ranked_plain_query_rate": 0.0125, "reward/invalid_output_rate": 0.0125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.9625, "reward/mean_mrr": 0.9875, "reward/mean_ndcg_at_10": 0.89765625, "reward/mean_r_ndcg": 0.78515625, "reward/mean_r_ndcg_cutoff": 5.93125, "reward/mean_recall": 0.2291015625, "reward/mean_recall_at_10": 0.048193359375, "reward/mean_recall_at_100": 0.2291015625, "reward/mean_retrieval_combined_bounded": 0.02689208984375, "reward/mean_retrieval_combined_raw": 0.04718598425388336, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 932.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2977155208587646, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.2984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6279805183410643, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15306090712547302, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.211408233642578, "sampling/raw_vllm_sampling_logp_difference/mean": 0.29469371438026426, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 935, "step_time": 13.870452308608218 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 74.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.403125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.49520449712872505, "epoch": 3.7154150197628457, "frac_reward_zero_std": 0.0, "grad_norm": 1.5930891036987305, "learning_rate": 2e-05, "loss": 0.0421875, "num_tokens": 56145352.0, "reward": 0.999127197265625, "reward/completion_length_max": 74.0, "reward/completion_length_mean": 32.45, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 35.4, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99921875, "reward/mean_colbert": 0.965625, "reward/mean_mrr": 0.990625, "reward/mean_ndcg_at_10": 0.8953125, "reward/mean_r_ndcg": 0.7734375, "reward/mean_r_ndcg_cutoff": 6.19375, "reward/mean_recall": 0.2431640625, "reward/mean_recall_at_10": 0.048193359375, "reward/mean_recall_at_100": 0.2431640625, "reward/mean_retrieval_combined_bounded": 0.0209381103515625, "reward/mean_retrieval_combined_raw": 0.034081704542040826, "reward/output_tag_rate": 0.99375, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 936.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.36252824664115907, "rewards/retrieval_reward/mean": 0.99921875, "rewards/retrieval_reward/std": 0.362890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.536246657371521, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1595137506723404, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.148604774475098, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2857981979846954, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 940, "step_time": 15.526670449774247 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.003125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.5006186723709106, "epoch": 3.7351778656126484, "frac_reward_zero_std": 0.0, "grad_norm": 1.6089212894439697, "learning_rate": 2e-05, "loss": 0.00054931640625, "num_tokens": 56443820.0, "reward": 1.020989990234375, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 31.975, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 35.4, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.0125, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.9609375, "reward/mean_mrr": 0.9875, "reward/mean_ndcg_at_10": 0.88203125, "reward/mean_r_ndcg": 0.72734375, "reward/mean_r_ndcg_cutoff": 5.08125, "reward/mean_recall": 0.231640625, "reward/mean_recall_at_10": 0.0486328125, "reward/mean_recall_at_100": 0.231640625, "reward/mean_retrieval_combined_bounded": 0.03465576171875, "reward/mean_retrieval_combined_raw": 0.05878629684448242, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 942.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.29144343733787537, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.29140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.9147600650787355, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16213254928588866, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.52092113494873, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2655880630016327, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 945, "step_time": 13.884901843778788 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.78125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.4, "completions/min_terminated_length": 0.0, "entropy": 0.436458470672369, "epoch": 3.7549407114624507, "frac_reward_zero_std": 0.0, "grad_norm": 1.531647801399231, "learning_rate": 2e-05, "loss": -0.00091552734375, "num_tokens": 56743252.0, "reward": 0.99420166015625, "reward/completion_length_max": 39.2, "reward/completion_length_mean": 32.8, "reward/completion_length_min": 27.4, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.2, "reward/duplicate_query_rate": 0.03046875, "reward/duplicate_ranked_plain_query_rate": 0.03046875, "reward/invalid_output_rate": 0.03125, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99375, "reward/mean_colbert": 0.94609375, "reward/mean_mrr": 0.96875, "reward/mean_ndcg_at_10": 0.8578125, "reward/mean_r_ndcg": 0.7171875, "reward/mean_r_ndcg_cutoff": 5.70625, "reward/mean_recall": 0.2267578125, "reward/mean_recall_at_10": 0.0474609375, "reward/mean_recall_at_100": 0.2267578125, "reward/mean_retrieval_combined_bounded": 0.02578125, "reward/mean_retrieval_combined_raw": 0.05863921642303467, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.96875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 946.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.33000574111938474, "rewards/retrieval_reward/mean": 0.99375, "rewards/retrieval_reward/std": 0.33046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5987220764160157, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.14926509857177733, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.640642356872558, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26899499297142027, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 950, "step_time": 13.908044544188305 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.7265625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.49787865579128265, "epoch": 3.758893280632411, "frac_reward_zero_std": 0.0, "num_tokens": 56803022.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6712146930675383, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7212423698026877, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8936731107205624, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9973637961335676, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5846230875583481, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6438570905617761, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8466666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.32465047031598393, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3661778408314435, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5349397590361445, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8409638554216867, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6193142454479816, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6285913212519834, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7552870251813913, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.905575117370892, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5626746081217529, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.35299794946141044, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17243422888985033, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.30327895401810395, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 1.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3451.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.41100101142563267, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.44801262345694154, "retrieval_eval/nq/single_turn/recall_at_10": 0.6449338356032067, "retrieval_eval/nq/single_turn/recall_at_100": 0.8817009562445668, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6766813723720011, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7022465542242129, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8295555555555556, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9426666666666667, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.046661376953125, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 32.75, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.01171875, "reward/duplicate_ranked_plain_query_rate": 0.01171875, "reward/invalid_output_rate": 0.01171875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.046875, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.95703125, "reward/mean_r_ndcg": 0.85546875, "reward/mean_r_ndcg_cutoff": 7.03125, "reward/mean_recall": 0.1748046875, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.1748046875, "reward/mean_retrieval_combined_bounded": 0.05859375, "reward/mean_retrieval_combined_raw": 0.07666015625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 952.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2701980471611023, "rewards/retrieval_reward/mean": 1.046875, "rewards/retrieval_reward/std": 0.271484375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5458109378814697, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.0781036764383316, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.483573913574219, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3164430260658264, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 951, "step_time": 13.818123581062537 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.75, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.9296875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.75, "completions/min_terminated_length": 0.0, "entropy": 0.4523252295330167, "epoch": 3.774703557312253, "frac_reward_zero_std": 0.0, "grad_norm": 1.5321614742279053, "learning_rate": 2e-05, "loss": -0.0003082275390625, "num_tokens": 57042566.0, "reward": 1.0165061950683594, "reward/completion_length_max": 38.75, "reward/completion_length_mean": 32.9375, "reward/completion_length_min": 28.75, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 36.25, "reward/duplicate_query_rate": 0.0087890625, "reward/duplicate_ranked_plain_query_rate": 0.0087890625, "reward/invalid_output_rate": 0.0087890625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.017578125, "reward/mean_colbert": 0.970703125, "reward/mean_mrr": 0.9912109375, "reward/mean_ndcg_at_10": 0.8994140625, "reward/mean_r_ndcg": 0.810546875, "reward/mean_r_ndcg_cutoff": 6.1796875, "reward/mean_recall": 0.23583984375, "reward/mean_recall_at_10": 0.04913330078125, "reward/mean_recall_at_100": 0.23583984375, "reward/mean_retrieval_combined_bounded": 0.025299072265625, "reward/mean_retrieval_combined_raw": 0.03863811492919922, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9912109375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9912109375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 952.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.30785226076841354, "rewards/retrieval_reward/mean": 1.017578125, "rewards/retrieval_reward/std": 0.30859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6993407011032104, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.14711210131645203, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.607306718826294, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26093464717268944, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 955, "step_time": 13.970679315534653 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.7078125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.4438333809375763, "epoch": 3.794466403162055, "frac_reward_zero_std": 0.0, "grad_norm": 1.9365594387054443, "learning_rate": 2e-05, "loss": -0.000634765625, "num_tokens": 57341008.0, "reward": 1.0133056640625, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 31.7, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01328125, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.87890625, "reward/mean_r_ndcg": 0.7828125, "reward/mean_r_ndcg_cutoff": 6.25, "reward/mean_recall": 0.2388671875, "reward/mean_recall_at_10": 0.048486328125, "reward/mean_recall_at_100": 0.2388671875, "reward/mean_retrieval_combined_bounded": 0.017254638671875, "reward/mean_retrieval_combined_raw": 0.021491193771362306, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 957.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.2892317295074463, "rewards/retrieval_reward/mean": 1.01328125, "rewards/retrieval_reward/std": 0.28984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.758868360519409, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13654221445322037, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.455883026123047, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2986262857913971, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 960, "step_time": 13.985458136012312 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.99453125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.4351130001246929, "epoch": 3.8142292490118574, "frac_reward_zero_std": 0.0, "grad_norm": 1.6909070014953613, "learning_rate": 2e-05, "loss": 0.0001068115234375, "num_tokens": 57639497.0, "reward": 1.012481689453125, "reward/completion_length_max": 37.2, "reward/completion_length_mean": 31.975, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01328125, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.88359375, "reward/mean_r_ndcg": 0.7375, "reward/mean_r_ndcg_cutoff": 6.025, "reward/mean_recall": 0.201171875, "reward/mean_recall_at_10": 0.047216796875, "reward/mean_recall_at_100": 0.201171875, "reward/mean_retrieval_combined_bounded": 0.0188507080078125, "reward/mean_retrieval_combined_raw": 0.037009131908416745, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 961.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28577411770820615, "rewards/retrieval_reward/mean": 1.01328125, "rewards/retrieval_reward/std": 0.2859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6999792575836183, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1401946008205414, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.370611572265625, "sampling/raw_vllm_sampling_logp_difference/mean": 0.30390815138816835, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 965, "step_time": 13.83813714144053 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.98203125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.40493144020438193, "epoch": 3.83399209486166, "frac_reward_zero_std": 0.0, "grad_norm": 1.5514439344406128, "learning_rate": 2e-05, "loss": -0.00299072265625, "num_tokens": 57939090.0, "reward": 1.0272674560546875, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 33.0, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.6, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.028125, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.88984375, "reward/mean_r_ndcg": 0.7828125, "reward/mean_r_ndcg_cutoff": 6.66875, "reward/mean_recall": 0.224609375, "reward/mean_recall_at_10": 0.047802734375, "reward/mean_recall_at_100": 0.224609375, "reward/mean_retrieval_combined_bounded": 0.0335205078125, "reward/mean_retrieval_combined_raw": 0.03346912264823913, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 967.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2903991758823395, "rewards/retrieval_reward/mean": 1.028125, "rewards/retrieval_reward/std": 0.291015625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8770843982696532, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15901175141334534, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.385943412780762, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2744755566120148, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 970, "step_time": 13.924563870613929 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 43.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.28203125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.4764732845127583, "epoch": 3.8537549407114624, "frac_reward_zero_std": 0.0, "grad_norm": 1.954274296760559, "learning_rate": 2e-05, "loss": 0.004052734375, "num_tokens": 58239195.0, "reward": 1.0346160888671876, "reward/completion_length_max": 43.2, "reward/completion_length_mean": 33.3, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 37.2, "reward/duplicate_query_rate": 0.0109375, "reward/duplicate_ranked_plain_query_rate": 0.0109375, "reward/invalid_output_rate": 0.0234375, "reward/length_out_of_range_rate": 0.0125, "reward/low_length_query_rate": 0.0125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.034375, "reward/mean_colbert": 0.953125, "reward/mean_mrr": 0.9765625, "reward/mean_ndcg_at_10": 0.8703125, "reward/mean_r_ndcg": 0.7484375, "reward/mean_r_ndcg_cutoff": 5.675, "reward/mean_recall": 0.2560546875, "reward/mean_recall_at_10": 0.0478515625, "reward/mean_recall_at_100": 0.2560546875, "reward/mean_retrieval_combined_bounded": 0.06416015625, "reward/mean_retrieval_combined_raw": 0.10732421148568391, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 972.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3211482346057892, "rewards/retrieval_reward/mean": 1.034375, "rewards/retrieval_reward/std": 0.321875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.604112148284912, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15700757503509521, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.497488117218017, "sampling/raw_vllm_sampling_logp_difference/mean": 0.28128454387187957, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 975, "step_time": 14.132182060810738 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 35.06640625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.6, "completions/min_terminated_length": 0.0, "entropy": 0.528533460944891, "epoch": 3.8735177865612647, "frac_reward_zero_std": 0.0, "grad_norm": 1.4046789407730103, "learning_rate": 2e-05, "loss": 0.00419921875, "num_tokens": 58541456.0, "reward": 1.0306915283203124, "reward/completion_length_max": 41.0, "reward/completion_length_mean": 35.1, "reward/completion_length_min": 30.6, "reward/completion_length_p90": 37.4, "reward/completion_length_p95": 38.4, "reward/duplicate_query_rate": 0.01875, "reward/duplicate_ranked_plain_query_rate": 0.01875, "reward/invalid_output_rate": 0.0328125, "reward/length_out_of_range_rate": 0.0140625, "reward/low_length_query_rate": 0.0140625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03046875, "reward/mean_colbert": 0.94453125, "reward/mean_mrr": 0.9671875, "reward/mean_ndcg_at_10": 0.90390625, "reward/mean_r_ndcg": 0.81953125, "reward/mean_r_ndcg_cutoff": 6.6625, "reward/mean_recall": 0.215234375, "reward/mean_recall_at_10": 0.04794921875, "reward/mean_recall_at_100": 0.215234375, "reward/mean_retrieval_combined_bounded": 0.070654296875, "reward/mean_retrieval_combined_raw": 0.136865234375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9671875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9671875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 976.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0140625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3359099566936493, "rewards/retrieval_reward/mean": 1.03046875, "rewards/retrieval_reward/std": 0.33671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.768368101119995, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13247129917144776, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.391357231140137, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2831528216600418, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 980, "step_time": 14.069895428430755 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 43.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.17265625, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.6, "completions/min_terminated_length": 0.0, "entropy": 0.44193080961704256, "epoch": 3.8932806324110674, "frac_reward_zero_std": 0.0, "grad_norm": 1.7603659629821777, "learning_rate": 2e-05, "loss": 0.007861328125, "num_tokens": 58845069.0, "reward": 0.99248046875, "reward/completion_length_max": 43.2, "reward/completion_length_mean": 36.2, "reward/completion_length_min": 31.6, "reward/completion_length_p90": 39.4, "reward/completion_length_p95": 40.4, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.06015625, "reward/length_out_of_range_rate": 0.05, "reward/low_length_query_rate": 0.05, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99296875, "reward/mean_colbert": 0.91796875, "reward/mean_mrr": 0.93984375, "reward/mean_ndcg_at_10": 0.84375, "reward/mean_r_ndcg": 0.7234375, "reward/mean_r_ndcg_cutoff": 6.5875, "reward/mean_recall": 0.207421875, "reward/mean_recall_at_10": 0.04541015625, "reward/mean_recall_at_100": 0.207421875, "reward/mean_retrieval_combined_bounded": 0.0775146484375, "reward/mean_retrieval_combined_raw": 0.1722655177116394, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.93984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.93984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 982.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.05, "reward/too_short_query_rate": 0.0, "reward_std": 0.42169383764266966, "rewards/retrieval_reward/mean": 0.99296875, "rewards/retrieval_reward/std": 0.421875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4118354320526123, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12806710079312325, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.52247142791748, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26659472584724425, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 985, "step_time": 14.111889459786472 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 60.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 35.4359375, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.2, "completions/min_terminated_length": 0.0, "entropy": 0.48794670552015307, "epoch": 3.9130434782608696, "frac_reward_zero_std": 0.0, "grad_norm": 1.3940097093582153, "learning_rate": 2e-05, "loss": 0.014453125, "num_tokens": 59148507.0, "reward": 1.0016693115234374, "reward/completion_length_max": 60.0, "reward/completion_length_mean": 35.45, "reward/completion_length_min": 31.2, "reward/completion_length_p90": 38.2, "reward/completion_length_p95": 39.4, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.03984375, "reward/length_out_of_range_rate": 0.03203125, "reward/low_length_query_rate": 0.03203125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00234375, "reward/mean_colbert": 0.9359375, "reward/mean_mrr": 0.96015625, "reward/mean_ndcg_at_10": 0.8703125, "reward/mean_r_ndcg": 0.784375, "reward/mean_r_ndcg_cutoff": 6.08125, "reward/mean_recall": 0.2150390625, "reward/mean_recall_at_10": 0.046728515625, "reward/mean_recall_at_100": 0.2150390625, "reward/mean_retrieval_combined_bounded": 0.057470703125, "reward/mean_retrieval_combined_raw": 0.1120849609375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.96015625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96015625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 986.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.03203125, "reward/too_short_query_rate": 0.0, "reward_std": 0.37500653266906736, "rewards/retrieval_reward/mean": 1.00234375, "rewards/retrieval_reward/std": 0.375390625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.484045720100403, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.14084534645080565, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.408407402038574, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2704949349164963, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 990, "step_time": 14.950635170401075 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.13984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.8, "completions/min_terminated_length": 0.0, "entropy": 0.4601341255009174, "epoch": 3.932806324110672, "frac_reward_zero_std": 0.0, "grad_norm": 1.8852968215942383, "learning_rate": 2e-05, "loss": 0.00186767578125, "num_tokens": 59449934.0, "reward": 1.0241851806640625, "reward/completion_length_max": 40.6, "reward/completion_length_mean": 34.2, "reward/completion_length_min": 29.8, "reward/completion_length_p90": 36.8, "reward/completion_length_p95": 37.6, "reward/duplicate_query_rate": 0.015625, "reward/duplicate_ranked_plain_query_rate": 0.015625, "reward/invalid_output_rate": 0.02109375, "reward/length_out_of_range_rate": 0.00546875, "reward/low_length_query_rate": 0.00546875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.9546875, "reward/mean_mrr": 0.978125, "reward/mean_ndcg_at_10": 0.840625, "reward/mean_r_ndcg": 0.67265625, "reward/mean_r_ndcg_cutoff": 5.5125, "reward/mean_recall": 0.221484375, "reward/mean_recall_at_10": 0.047265625, "reward/mean_recall_at_100": 0.221484375, "reward/mean_retrieval_combined_bounded": 0.0481201171875, "reward/mean_retrieval_combined_raw": 0.08950292468070983, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 992.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00546875, "reward/too_short_query_rate": 0.0, "reward_std": 0.30106961727142334, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.301171875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.641676425933838, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12162368446588516, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.555004119873047, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2865209460258484, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 995, "step_time": 14.058283553190995 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.6875, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.436977206915617, "epoch": 3.9525691699604746, "frac_reward_zero_std": 0.0, "grad_norm": 1.8984403610229492, "learning_rate": 2e-05, "loss": -0.00146484375, "num_tokens": 59748990.0, "reward": 1.013677978515625, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 32.675, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 36.2, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.003125, "reward/malformed_ranked_plain_query_rate": 0.003125, "reward/mean": 1.0140625, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.90078125, "reward/mean_r_ndcg": 0.7609375, "reward/mean_r_ndcg_cutoff": 6.01875, "reward/mean_recall": 0.2298828125, "reward/mean_recall_at_10": 0.049267578125, "reward/mean_recall_at_100": 0.2298828125, "reward/mean_retrieval_combined_bounded": 0.023699188232421876, "reward/mean_retrieval_combined_raw": 0.03247807621955871, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 997.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3051347672939301, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.305859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8285796642303467, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16218532621860504, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.1857759475708, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2987235724925995, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1000, "step_time": 13.89490273144329 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.46484375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.3857571743428707, "epoch": 3.9565217391304346, "frac_reward_zero_std": 0.0, "num_tokens": 59808693.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6605863002476087, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7158073037039946, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9024604569420035, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9947275922671354, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5879026269233043, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6448910266754835, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.84, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.32430952955521997, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.36139288683268994, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5228915662650603, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8433734939759037, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6224979582169625, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6252844698144723, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.73749275958783, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9078931924882628, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5625146581812419, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3455150846150288, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17343688301420537, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.30412442985351157, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 1.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3451.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.40825450424814774, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4450588180816203, "retrieval_eval/nq/single_turn/recall_at_10": 0.6456341157152516, "retrieval_eval/nq/single_turn/recall_at_100": 0.8815560707041438, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6664879224259618, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6989513709451497, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8378888888888889, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9433333333333334, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.02777099609375, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 32.5, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.98046875, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.86328125, "reward/mean_r_ndcg": 0.80078125, "reward/mean_r_ndcg_cutoff": 8.125, "reward/mean_recall": 0.2734375, "reward/mean_recall_at_10": 0.0498046875, "reward/mean_recall_at_100": 0.2734375, "reward/mean_retrieval_combined_bounded": 0.0277099609375, "reward/mean_retrieval_combined_raw": 8.493661880493164e-07, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1000.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2951073944568634, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.294921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.938840866088867, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20201268792152405, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.12500286102295, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27064988017082214, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1001, "step_time": 13.855445173860062 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.5419921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.25, "completions/min_terminated_length": 0.0, "entropy": 0.4208053257316351, "epoch": 3.972332015810277, "frac_reward_zero_std": 0.0, "grad_norm": 1.6451414823532104, "learning_rate": 2e-05, "loss": -0.001611328125, "num_tokens": 60046752.0, "reward": 1.0057716369628906, "reward/completion_length_max": 36.25, "reward/completion_length_mean": 31.53125, "reward/completion_length_min": 28.25, "reward/completion_length_p90": 33.75, "reward/completion_length_p95": 34.25, "reward/duplicate_query_rate": 0.0009765625, "reward/duplicate_ranked_plain_query_rate": 0.0009765625, "reward/invalid_output_rate": 0.0009765625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.005859375, "reward/mean_colbert": 0.9755859375, "reward/mean_mrr": 0.9990234375, "reward/mean_ndcg_at_10": 0.892578125, "reward/mean_r_ndcg": 0.734375, "reward/mean_r_ndcg_cutoff": 7.0234375, "reward/mean_recall": 0.23828125, "reward/mean_recall_at_10": 0.04864501953125, "reward/mean_recall_at_100": 0.23828125, "reward/mean_retrieval_combined_bounded": 0.006626605987548828, "reward/mean_retrieval_combined_raw": 0.00799005851149559, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9990234375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9990234375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1002.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2792602404952049, "rewards/retrieval_reward/mean": 1.005859375, "rewards/retrieval_reward/std": 0.279296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6997573375701904, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1520203836262226, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.26923656463623, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2583096958696842, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1005, "step_time": 13.894314096527523 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.0578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.8, "completions/min_terminated_length": 0.0, "entropy": 0.37829427048563957, "epoch": 3.992094861660079, "frac_reward_zero_std": 0.0, "grad_norm": 2.3325271606445312, "learning_rate": 2e-05, "loss": -0.001177978515625, "num_tokens": 60342410.0, "reward": 1.0222991943359374, "reward/completion_length_max": 35.4, "reward/completion_length_mean": 30.075, "reward/completion_length_min": 26.8, "reward/completion_length_p90": 32.4, "reward/completion_length_p95": 33.4, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.975, "reward/mean_mrr": 0.99765625, "reward/mean_ndcg_at_10": 0.86328125, "reward/mean_r_ndcg": 0.7390625, "reward/mean_r_ndcg_cutoff": 6.44375, "reward/mean_recall": 0.2275390625, "reward/mean_recall_at_10": 0.047998046875, "reward/mean_recall_at_100": 0.2275390625, "reward/mean_retrieval_combined_bounded": 0.0230255126953125, "reward/mean_retrieval_combined_raw": 0.00632790606468916, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1007.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2819949209690094, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.28203125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7184918880462647, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1896409958600998, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.836929321289062, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25637357831001284, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1010, "step_time": 13.870770989195444 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.7625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.2, "completions/min_terminated_length": 0.0, "entropy": 0.31973436996340754, "epoch": 4.011857707509882, "frac_reward_zero_std": 0.0, "grad_norm": 1.8341524600982666, "learning_rate": 2e-05, "loss": 8.96453857421875e-05, "num_tokens": 60636858.0, "reward": 1.0155059814453125, "reward/completion_length_max": 36.0, "reward/completion_length_mean": 28.75, "reward/completion_length_min": 25.2, "reward/completion_length_p90": 31.6, "reward/completion_length_p95": 32.4, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00234375, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01484375, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.99765625, "reward/mean_ndcg_at_10": 0.89453125, "reward/mean_r_ndcg": 0.76328125, "reward/mean_r_ndcg_cutoff": 6.1125, "reward/mean_recall": 0.2328125, "reward/mean_recall_at_10": 0.04853515625, "reward/mean_recall_at_100": 0.2328125, "reward/mean_retrieval_combined_bounded": 0.018670654296875, "reward/mean_retrieval_combined_raw": 0.018996618688106537, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1012.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0015625, "reward_std": 0.26750283539295194, "rewards/retrieval_reward/mean": 1.01484375, "rewards/retrieval_reward/std": 0.26796875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.794317293167114, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1743537575006485, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.649489974975586, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26267164647579194, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1015, "step_time": 13.756811181805096 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.22578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.2, "completions/min_terminated_length": 0.0, "entropy": 0.3437099665403366, "epoch": 4.031620553359684, "frac_reward_zero_std": 0.0, "grad_norm": 2.1455986499786377, "learning_rate": 2e-05, "loss": -0.001513671875, "num_tokens": 60930075.0, "reward": 1.0125335693359374, "reward/completion_length_max": 33.8, "reward/completion_length_mean": 28.2, "reward/completion_length_min": 25.2, "reward/completion_length_p90": 30.4, "reward/completion_length_p95": 31.0, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.01171875, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.00625, "reward/malformed_ranked_plain_query_rate": 0.00625, "reward/mean": 1.01015625, "reward/mean_colbert": 0.9640625, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.878125, "reward/mean_r_ndcg": 0.7484375, "reward/mean_r_ndcg_cutoff": 6.03125, "reward/mean_recall": 0.23671875, "reward/mean_recall_at_10": 0.0478515625, "reward/mean_recall_at_100": 0.23671875, "reward/mean_retrieval_combined_bounded": 0.03087158203125, "reward/mean_retrieval_combined_raw": 0.04291331768035889, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1016.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.32253423929214475, "rewards/retrieval_reward/mean": 1.01015625, "rewards/retrieval_reward/std": 0.323828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.927118921279907, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21525904834270476, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.876254463195801, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2723113030195236, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1020, "step_time": 13.62199471766362 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 51.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 27.646875, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.8, "completions/min_terminated_length": 0.0, "entropy": 0.33540762811899183, "epoch": 4.051383399209486, "frac_reward_zero_std": 0.0, "grad_norm": 2.0275609493255615, "learning_rate": 2e-05, "loss": 0.00927734375, "num_tokens": 61222935.0, "reward": 1.010711669921875, "reward/completion_length_max": 51.6, "reward/completion_length_mean": 27.65, "reward/completion_length_min": 24.8, "reward/completion_length_p90": 29.8, "reward/completion_length_p95": 30.4, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.0109375, "reward/length_out_of_range_rate": 0.00703125, "reward/low_length_query_rate": 0.00703125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0125, "reward/mean_colbert": 0.96328125, "reward/mean_mrr": 0.9890625, "reward/mean_ndcg_at_10": 0.896875, "reward/mean_r_ndcg": 0.75078125, "reward/mean_r_ndcg_cutoff": 5.9125, "reward/mean_recall": 0.2275390625, "reward/mean_recall_at_10": 0.04873046875, "reward/mean_recall_at_100": 0.2275390625, "reward/mean_retrieval_combined_bounded": 0.0252685546875, "reward/mean_retrieval_combined_raw": 0.04985349448397756, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1022.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00703125, "reward_std": 0.30795949697494507, "rewards/retrieval_reward/mean": 1.0109375, "rewards/retrieval_reward/std": 0.308203125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.636515665054321, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.19899725615978242, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.070487022399902, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26598802804946897, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1025, "step_time": 14.446831820590887 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.0796875, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.0, "completions/min_terminated_length": 0.0, "entropy": 0.3264370009303093, "epoch": 4.071146245059288, "frac_reward_zero_std": 0.0, "grad_norm": 2.460200786590576, "learning_rate": 2e-05, "loss": -0.0007720947265625, "num_tokens": 61516253.0, "reward": 1.0159515380859374, "reward/completion_length_max": 32.8, "reward/completion_length_mean": 28.075, "reward/completion_length_min": 25.0, "reward/completion_length_p90": 30.6, "reward/completion_length_p95": 31.0, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.01015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.965625, "reward/mean_mrr": 0.98984375, "reward/mean_ndcg_at_10": 0.903125, "reward/mean_r_ndcg": 0.76875, "reward/mean_r_ndcg_cutoff": 6.14375, "reward/mean_recall": 0.2123046875, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.2123046875, "reward/mean_retrieval_combined_bounded": 0.02607879638671875, "reward/mean_retrieval_combined_raw": 0.04008878222666681, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1025.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2861594557762146, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.286328125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6505412101745605, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18710613250732422, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.556319427490234, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26035469174385073, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1030, "step_time": 13.548359219590202 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.48515625, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.6, "completions/min_terminated_length": 0.0, "entropy": 0.29713328257203103, "epoch": 4.090909090909091, "frac_reward_zero_std": 0.0, "grad_norm": 3.391084909439087, "learning_rate": 2e-05, "loss": 0.000689697265625, "num_tokens": 61810602.0, "reward": 0.997601318359375, "reward/completion_length_max": 33.6, "reward/completion_length_mean": 28.475, "reward/completion_length_min": 24.6, "reward/completion_length_p90": 30.8, "reward/completion_length_p95": 31.6, "reward/duplicate_query_rate": 0.009375, "reward/duplicate_ranked_plain_query_rate": 0.009375, "reward/invalid_output_rate": 0.01328125, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.9984375, "reward/mean_colbert": 0.96328125, "reward/mean_mrr": 0.98671875, "reward/mean_ndcg_at_10": 0.896875, "reward/mean_r_ndcg": 0.81171875, "reward/mean_r_ndcg_cutoff": 7.0125, "reward/mean_recall": 0.2283203125, "reward/mean_recall_at_10": 0.048486328125, "reward/mean_recall_at_100": 0.2283203125, "reward/mean_retrieval_combined_bounded": 0.012786865234375, "reward/mean_retrieval_combined_raw": -0.09546875953674316, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98671875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98671875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1032.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00390625, "reward_std": 0.3240490972995758, "rewards/retrieval_reward/mean": 0.9984375, "rewards/retrieval_reward/std": 0.324609375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5538339614868164, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18471306562423706, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.407873916625977, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24105999171733855, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1035, "step_time": 13.592694704013411 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 27.3703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.2, "completions/min_terminated_length": 0.0, "entropy": 0.28186209239065646, "epoch": 4.1106719367588935, "frac_reward_zero_std": 0.0, "grad_norm": 2.2651584148406982, "learning_rate": 2e-05, "loss": -0.0039794921875, "num_tokens": 62102852.0, "reward": 1.02066650390625, "reward/completion_length_max": 34.2, "reward/completion_length_mean": 27.4, "reward/completion_length_min": 24.2, "reward/completion_length_p90": 30.0, "reward/completion_length_p95": 30.6, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0046875, "reward/low_length_query_rate": 0.0046875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.89609375, "reward/mean_r_ndcg": 0.73984375, "reward/mean_r_ndcg_cutoff": 5.675, "reward/mean_recall": 0.2400390625, "reward/mean_recall_at_10": 0.048876953125, "reward/mean_recall_at_100": 0.2400390625, "reward/mean_retrieval_combined_bounded": 0.0292724609375, "reward/mean_retrieval_combined_raw": 0.04326171875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1038.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.00390625, "reward_std": 0.2877415418624878, "rewards/retrieval_reward/mean": 1.02109375, "rewards/retrieval_reward/std": 0.287890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.803418493270874, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2375752180814743, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.32086353302002, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22525788843631744, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1040, "step_time": 13.61557023039786 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.4375, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.6, "completions/min_terminated_length": 0.0, "entropy": 0.3062067836523056, "epoch": 4.130434782608695, "frac_reward_zero_std": 0.0, "grad_norm": 1.6554348468780518, "learning_rate": 2e-05, "loss": -0.00013580322265625, "num_tokens": 62396564.0, "reward": 1.0106689453125, "reward/completion_length_max": 32.6, "reward/completion_length_mean": 28.4, "reward/completion_length_min": 25.6, "reward/completion_length_p90": 30.4, "reward/completion_length_p95": 31.2, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0109375, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.8640625, "reward/mean_r_ndcg": 0.73203125, "reward/mean_r_ndcg_cutoff": 5.78125, "reward/mean_recall": 0.216015625, "reward/mean_recall_at_10": 0.048291015625, "reward/mean_recall_at_100": 0.216015625, "reward/mean_retrieval_combined_bounded": 0.014111328125, "reward/mean_retrieval_combined_raw": 0.02158176079392433, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1040.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.2654407888650894, "rewards/retrieval_reward/mean": 1.0109375, "rewards/retrieval_reward/std": 0.2662109375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.798891878128052, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20737117528915405, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.494160652160645, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24149932265281676, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1045, "step_time": 13.662523408187553 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 51.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.84765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.0, "completions/min_terminated_length": 0.0, "entropy": 0.3505411364138126, "epoch": 4.150197628458498, "frac_reward_zero_std": 0.0, "grad_norm": 2.0948257446289062, "learning_rate": 2e-05, "loss": 0.011865234375, "num_tokens": 62690609.0, "reward": 1.038397216796875, "reward/completion_length_max": 51.4, "reward/completion_length_mean": 28.85, "reward/completion_length_min": 26.0, "reward/completion_length_p90": 30.4, "reward/completion_length_p95": 31.0, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.00546875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0390625, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.99453125, "reward/mean_ndcg_at_10": 0.91171875, "reward/mean_r_ndcg": 0.79375, "reward/mean_r_ndcg_cutoff": 6.0125, "reward/mean_recall": 0.21875, "reward/mean_recall_at_10": 0.048486328125, "reward/mean_recall_at_100": 0.21875, "reward/mean_retrieval_combined_bounded": 0.04541015625, "reward/mean_retrieval_combined_raw": 0.07338846027851105, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.99453125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99453125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1048.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28507023453712466, "rewards/retrieval_reward/mean": 1.0390625, "rewards/retrieval_reward/std": 0.28515625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.798997402191162, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.22128461599349974, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.261692810058594, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2547531247138977, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1050, "step_time": 14.523345715017058 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.4921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.0, "completions/min_terminated_length": 0.0, "entropy": 0.3374338373541832, "epoch": 4.154150197628459, "frac_reward_zero_std": 0.0, "num_tokens": 62749551.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6698172139140618, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7209653786027397, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8963093145869947, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9947275922671354, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5501447300681386, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6090215394647424, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8133333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3114320598100694, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3444024067568436, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.4963855421686747, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8409638554216867, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6109665122764124, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.623397354989981, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7553624778976892, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9060487927565392, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5568447523881271, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3515033607906475, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17582243276346868, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3051587320583956, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 0.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3452.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.40043644619587326, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4362710708945851, "retrieval_eval/nq/single_turn/recall_at_10": 0.6363943607570491, "retrieval_eval/nq/single_turn/recall_at_100": 0.876038045577443, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6591710534467896, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6936257946286295, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8323333333333333, "retrieval_eval/scifact/single_turn/recall_at_100": 0.938888888888889, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0175018310546875, "reward/completion_length_max": 37.0, "reward/completion_length_mean": 29.5, "reward/completion_length_min": 26.0, "reward/completion_length_p90": 31.0, "reward/completion_length_p95": 32.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.98046875, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.94921875, "reward/mean_r_ndcg": 0.78515625, "reward/mean_r_ndcg_cutoff": 6.875, "reward/mean_recall": 0.2265625, "reward/mean_recall_at_10": 0.049560546875, "reward/mean_recall_at_100": 0.2265625, "reward/mean_retrieval_combined_bounded": 0.0177001953125, "reward/mean_retrieval_combined_raw": 1.4781951904296875e-05, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1048.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2844034731388092, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.28515625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.9703633785247803, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16297240555286407, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.954564094543457, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23710264265537262, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1051, "step_time": 14.04109315003734 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.1982421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.3645938094705343, "epoch": 4.169960474308301, "frac_reward_zero_std": 0.0, "grad_norm": 2.618903160095215, "learning_rate": 2e-05, "loss": 0.0006988525390625, "num_tokens": 62986522.0, "reward": 1.0343780517578125, "reward/completion_length_max": 35.0, "reward/completion_length_mean": 30.1875, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 32.5, "reward/duplicate_query_rate": 0.01171875, "reward/duplicate_ranked_plain_query_rate": 0.01171875, "reward/invalid_output_rate": 0.01171875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03515625, "reward/mean_colbert": 0.9638671875, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.8974609375, "reward/mean_r_ndcg": 0.8017578125, "reward/mean_r_ndcg_cutoff": 6.3203125, "reward/mean_recall": 0.21826171875, "reward/mean_recall_at_10": 0.04827880859375, "reward/mean_recall_at_100": 0.21826171875, "reward/mean_retrieval_combined_bounded": 0.04608154296875, "reward/mean_retrieval_combined_raw": 0.052975716069340706, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1054.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28852319717407227, "rewards/retrieval_reward/mean": 1.03515625, "rewards/retrieval_reward/std": 0.28857421875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8554946184158325, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1937527060508728, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.096638202667236, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2554738372564316, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1055, "step_time": 13.877300274223671 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.1578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.3898549571633339, "epoch": 4.189723320158103, "frac_reward_zero_std": 0.0, "grad_norm": 2.5075876712799072, "learning_rate": 2e-05, "loss": -0.00255126953125, "num_tokens": 63282628.0, "reward": 1.0048065185546875, "reward/completion_length_max": 34.0, "reward/completion_length_mean": 30.175, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 32.2, "reward/completion_length_p95": 32.4, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00390625, "reward/mean_colbert": 0.9671875, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.91640625, "reward/mean_r_ndcg": 0.7703125, "reward/mean_r_ndcg_cutoff": 6.19375, "reward/mean_recall": 0.2279296875, "reward/mean_recall_at_10": 0.049072265625, "reward/mean_recall_at_100": 0.2279296875, "reward/mean_retrieval_combined_bounded": 0.0126708984375, "reward/mean_retrieval_combined_raw": 0.02230639010667801, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1056.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28781151175498965, "rewards/retrieval_reward/mean": 1.00390625, "rewards/retrieval_reward/std": 0.28828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.843500518798828, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18885937631130217, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.133648300170899, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25967885851860045, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1060, "step_time": 13.825330165191554 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.084375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.8, "completions/min_terminated_length": 0.0, "entropy": 0.39967892989516257, "epoch": 4.209486166007905, "frac_reward_zero_std": 0.0, "grad_norm": 2.293459177017212, "learning_rate": 2e-05, "loss": -0.0009765625, "num_tokens": 63579696.0, "reward": 1.0394683837890626, "reward/completion_length_max": 37.2, "reward/completion_length_mean": 31.1, "reward/completion_length_min": 27.8, "reward/completion_length_p90": 33.8, "reward/completion_length_p95": 34.6, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.01015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0375, "reward/mean_colbert": 0.965625, "reward/mean_mrr": 0.98984375, "reward/mean_ndcg_at_10": 0.89609375, "reward/mean_r_ndcg": 0.78359375, "reward/mean_r_ndcg_cutoff": 6.00625, "reward/mean_recall": 0.2591796875, "reward/mean_recall_at_10": 0.0482421875, "reward/mean_recall_at_100": 0.2591796875, "reward/mean_retrieval_combined_bounded": 0.0496826171875, "reward/mean_retrieval_combined_raw": 0.1498391419649124, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1062.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.29870182275772095, "rewards/retrieval_reward/mean": 1.0375, "rewards/retrieval_reward/std": 0.298828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.871876573562622, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2504220813512802, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.828950500488281, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22670873403549194, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1065, "step_time": 13.887187765003182 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.17734375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.4484954759478569, "epoch": 4.229249011857707, "frac_reward_zero_std": 0.0, "grad_norm": 1.8929144144058228, "learning_rate": 2e-05, "loss": 5.078315734863281e-06, "num_tokens": 63876979.0, "reward": 1.0125823974609376, "reward/completion_length_max": 35.8, "reward/completion_length_mean": 31.175, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0125, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.90078125, "reward/mean_r_ndcg": 0.740625, "reward/mean_r_ndcg_cutoff": 5.21875, "reward/mean_recall": 0.2263671875, "reward/mean_recall_at_10": 0.04892578125, "reward/mean_recall_at_100": 0.2263671875, "reward/mean_retrieval_combined_bounded": 0.0134765625, "reward/mean_retrieval_combined_raw": 0.006395985186100006, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1067.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2850431501865387, "rewards/retrieval_reward/mean": 1.0125, "rewards/retrieval_reward/std": 0.285546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.755190706253052, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18091173768043517, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.737688159942627, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26607613265514374, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1070, "step_time": 13.808546335378196 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.659375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.8, "completions/min_terminated_length": 0.0, "entropy": 0.4758693613111973, "epoch": 4.24901185770751, "frac_reward_zero_std": 0.0, "grad_norm": 1.7891981601715088, "learning_rate": 2e-05, "loss": -0.0001220703125, "num_tokens": 64175039.0, "reward": 1.018359375, "reward/completion_length_max": 36.0, "reward/completion_length_mean": 31.675, "reward/completion_length_min": 28.8, "reward/completion_length_p90": 33.6, "reward/completion_length_p95": 34.2, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.87734375, "reward/mean_r_ndcg": 0.73125, "reward/mean_r_ndcg_cutoff": 5.55625, "reward/mean_recall": 0.2443359375, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.2443359375, "reward/mean_retrieval_combined_bounded": 0.021484375, "reward/mean_retrieval_combined_raw": 0.022419819235801698, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1072.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2672491788864136, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.267578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.774612045288086, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.17662070989608764, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.667560386657715, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2546390414237976, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1075, "step_time": 13.838848893961403 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.978125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.44590952470898626, "epoch": 4.2687747035573125, "frac_reward_zero_std": 0.0, "grad_norm": 1.6582297086715698, "learning_rate": 2e-05, "loss": -0.0002410888671875, "num_tokens": 64473507.0, "reward": 1.007965087890625, "reward/completion_length_max": 36.6, "reward/completion_length_mean": 32.0, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 34.2, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00859375, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.99765625, "reward/mean_ndcg_at_10": 0.86640625, "reward/mean_r_ndcg": 0.715625, "reward/mean_r_ndcg_cutoff": 5.6875, "reward/mean_recall": 0.21171875, "reward/mean_recall_at_10": 0.047705078125, "reward/mean_recall_at_100": 0.21171875, "reward/mean_retrieval_combined_bounded": 0.0095947265625, "reward/mean_retrieval_combined_raw": 0.01279399637132883, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1076.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2831577718257904, "rewards/retrieval_reward/mean": 1.00859375, "rewards/retrieval_reward/std": 0.28359375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8811583518981934, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20778534412384034, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.841645622253418, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23307847380638122, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1080, "step_time": 13.869533405604306 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.19140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.40628545805811883, "epoch": 4.288537549407114, "frac_reward_zero_std": 0.0, "grad_norm": 2.2269129753112793, "learning_rate": 2e-05, "loss": 0.000799560546875, "num_tokens": 64772312.0, "reward": 1.02589111328125, "reward/completion_length_max": 37.2, "reward/completion_length_mean": 32.175, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.9703125, "reward/mean_mrr": 0.9859375, "reward/mean_ndcg_at_10": 0.8890625, "reward/mean_r_ndcg": 0.79453125, "reward/mean_r_ndcg_cutoff": 6.525, "reward/mean_recall": 0.237109375, "reward/mean_recall_at_10": 0.04765625, "reward/mean_recall_at_100": 0.237109375, "reward/mean_retrieval_combined_bounded": 0.03050537109375, "reward/mean_retrieval_combined_raw": 0.029785561189055444, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1081.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2779103219509125, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.277734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7099648475646974, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.17861243784427644, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.614262008666993, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26839923560619355, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1085, "step_time": 13.834336786391214 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.7890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.4, "completions/min_terminated_length": 0.0, "entropy": 0.4663385070860386, "epoch": 4.308300395256917, "frac_reward_zero_std": 0.0, "grad_norm": 1.8514853715896606, "learning_rate": 2e-05, "loss": 0.0010009765625, "num_tokens": 65071658.0, "reward": 1.0268157958984374, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 32.8, "reward/completion_length_min": 29.4, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.0125, "reward/duplicate_ranked_plain_query_rate": 0.0125, "reward/invalid_output_rate": 0.0125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.9875, "reward/mean_ndcg_at_10": 0.90546875, "reward/mean_r_ndcg": 0.8046875, "reward/mean_r_ndcg_cutoff": 6.75625, "reward/mean_recall": 0.2208984375, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.2208984375, "reward/mean_retrieval_combined_bounded": 0.0392425537109375, "reward/mean_retrieval_combined_raw": 0.07963916063308715, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1088.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2970949411392212, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.297265625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.650440216064453, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15949435234069825, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.9430513381958, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27495476603507996, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1090, "step_time": 13.854086967802141 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 54.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.77109375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.8, "completions/min_terminated_length": 0.0, "entropy": 0.43958769366145134, "epoch": 4.32806324110672, "frac_reward_zero_std": 0.0, "grad_norm": 2.054915189743042, "learning_rate": 2e-05, "loss": 0.0193359375, "num_tokens": 65371333.0, "reward": 1.019818115234375, "reward/completion_length_max": 54.8, "reward/completion_length_mean": 32.8, "reward/completion_length_min": 28.8, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.96640625, "reward/mean_mrr": 0.990625, "reward/mean_ndcg_at_10": 0.89453125, "reward/mean_r_ndcg": 0.79296875, "reward/mean_r_ndcg_cutoff": 6.26875, "reward/mean_recall": 0.2328125, "reward/mean_recall_at_10": 0.047607421875, "reward/mean_recall_at_100": 0.2328125, "reward/mean_retrieval_combined_bounded": 0.03387451171875, "reward/mean_retrieval_combined_raw": 0.04472704827785492, "reward/output_tag_rate": 0.99765625, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1091.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.31178181767463686, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.31171875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.633297824859619, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1888771742582321, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.163387107849122, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25268023610115053, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1095, "step_time": 14.647561746148858 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.4046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.8, "completions/min_terminated_length": 0.0, "entropy": 0.39901404455304146, "epoch": 4.3478260869565215, "frac_reward_zero_std": 0.0, "grad_norm": 1.7021944522857666, "learning_rate": 2e-05, "loss": 0.0019775390625, "num_tokens": 65670475.0, "reward": 1.0078887939453125, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 32.35, "reward/completion_length_min": 28.8, "reward/completion_length_p90": 35.4, "reward/completion_length_p95": 36.2, "reward/duplicate_query_rate": 0.009375, "reward/duplicate_ranked_plain_query_rate": 0.009375, "reward/invalid_output_rate": 0.01015625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00703125, "reward/mean_colbert": 0.965625, "reward/mean_mrr": 0.98984375, "reward/mean_ndcg_at_10": 0.8953125, "reward/mean_r_ndcg": 0.76796875, "reward/mean_r_ndcg_cutoff": 6.5875, "reward/mean_recall": 0.2443359375, "reward/mean_recall_at_10": 0.04921875, "reward/mean_recall_at_100": 0.2443359375, "reward/mean_retrieval_combined_bounded": 0.0182373046875, "reward/mean_retrieval_combined_raw": 0.030174338817596437, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1096.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3149363934993744, "rewards/retrieval_reward/mean": 1.00703125, "rewards/retrieval_reward/std": 0.315234375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.798376512527466, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20009626746177672, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.070979690551757, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23773123621940612, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1100, "step_time": 13.85377520857146 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.7890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.4806681238114834, "epoch": 4.351778656126482, "frac_reward_zero_std": 0.0, "num_tokens": 65729781.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6545842667141358, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7057116475629683, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8840070298769771, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9956063268892794, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6065775123707634, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6591309520217604, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.84, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.32860418227814364, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3659358086243106, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5301204819277109, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8530120481927711, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6056367909117544, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6154596491613676, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7471114566184989, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9108609993293092, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5758044615813921, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3606374781225581, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.18158632383930295, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.318868024844499, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 3.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3449.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.4119001193455648, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.44901252719593765, "retrieval_eval/nq/single_turn/recall_at_10": 0.6497293901613994, "retrieval_eval/nq/single_turn/recall_at_100": 0.8845317483328501, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6664447946242047, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6968070204503762, "retrieval_eval/scifact/single_turn/recall_at_10": 0.835, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9426666666666667, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 0.9996337890625, "reward/completion_length_max": 34.0, "reward/completion_length_mean": 30.75, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0, "reward/mean_colbert": 0.98046875, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.94140625, "reward/mean_r_ndcg": 0.87890625, "reward/mean_r_ndcg_cutoff": 7.0, "reward/mean_recall": 0.259765625, "reward/mean_recall_at_10": 0.04931640625, "reward/mean_recall_at_100": 0.259765625, "reward/mean_retrieval_combined_bounded": -0.0002841949462890625, "reward/mean_retrieval_combined_raw": 1.125037670135498e-06, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1104.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2761264741420746, "rewards/retrieval_reward/mean": 1.0, "rewards/retrieval_reward/std": 0.27734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.607318639755249, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16312852501869202, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.664800643920898, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2814130485057831, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1101, "step_time": 13.649396501947194 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.8046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.75, "completions/min_terminated_length": 0.0, "entropy": 0.4314280226826668, "epoch": 4.367588932806324, "frac_reward_zero_std": 0.0, "grad_norm": 1.8739511966705322, "learning_rate": 2e-05, "loss": -0.001904296875, "num_tokens": 65968525.0, "reward": 0.9991111755371094, "reward/completion_length_max": 40.0, "reward/completion_length_mean": 31.78125, "reward/completion_length_min": 27.75, "reward/completion_length_p90": 34.25, "reward/completion_length_p95": 35.75, "reward/duplicate_query_rate": 0.0029296875, "reward/duplicate_ranked_plain_query_rate": 0.0029296875, "reward/invalid_output_rate": 0.0048828125, "reward/length_out_of_range_rate": 0.0009765625, "reward/low_length_query_rate": 0.0009765625, "reward/malformed_query_rate": 0.0009765625, "reward/malformed_ranked_plain_query_rate": 0.0009765625, "reward/mean": 1.0, "reward/mean_colbert": 0.970703125, "reward/mean_mrr": 0.9833984375, "reward/mean_ndcg_at_10": 0.876953125, "reward/mean_r_ndcg": 0.7216796875, "reward/mean_r_ndcg_cutoff": 6.6640625, "reward/mean_recall": 0.219482421875, "reward/mean_recall_at_10": 0.04864501953125, "reward/mean_recall_at_100": 0.219482421875, "reward/mean_retrieval_combined_bounded": 0.00547027587890625, "reward/mean_retrieval_combined_raw": -0.05981561541557312, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9951171875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9951171875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1104.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0009765625, "reward/too_short_query_rate": 0.0, "reward_std": 0.29757407307624817, "rewards/retrieval_reward/mean": 1.0, "rewards/retrieval_reward/std": 0.2978515625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7233000993728638, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.17948605492711067, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.086671113967896, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26466262340545654, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1105, "step_time": 14.055329370807158 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.9453125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.41370664909482, "epoch": 4.387351778656127, "frac_reward_zero_std": 0.0, "grad_norm": 1.4698537588119507, "learning_rate": 2e-05, "loss": -0.00185546875, "num_tokens": 66266247.0, "reward": 1.0339508056640625, "reward/completion_length_max": 37.2, "reward/completion_length_mean": 31.95, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.034375, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.88671875, "reward/mean_r_ndcg": 0.7796875, "reward/mean_r_ndcg_cutoff": 6.95, "reward/mean_recall": 0.2349609375, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.2349609375, "reward/mean_retrieval_combined_bounded": 0.0377197265625, "reward/mean_retrieval_combined_raw": 0.10610355362296105, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1105.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27544109225273133, "rewards/retrieval_reward/mean": 1.0328125, "rewards/retrieval_reward/std": 0.27578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.477767896652222, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20500050783157348, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.554014015197755, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2569777250289917, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1110, "step_time": 13.78837840121705 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.32578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.43442471250891684, "epoch": 4.407114624505929, "frac_reward_zero_std": 0.0, "grad_norm": 1.7108111381530762, "learning_rate": 2e-05, "loss": 0.000323486328125, "num_tokens": 66563752.0, "reward": 0.9959320068359375, "reward/completion_length_max": 36.4, "reward/completion_length_mean": 31.275, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 34.4, "reward/duplicate_query_rate": 0.025, "reward/duplicate_ranked_plain_query_rate": 0.025, "reward/invalid_output_rate": 0.025, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99609375, "reward/mean_colbert": 0.953125, "reward/mean_mrr": 0.975, "reward/mean_ndcg_at_10": 0.89453125, "reward/mean_r_ndcg": 0.8046875, "reward/mean_r_ndcg_cutoff": 6.9, "reward/mean_recall": 0.238671875, "reward/mean_recall_at_10": 0.048291015625, "reward/mean_recall_at_100": 0.238671875, "reward/mean_retrieval_combined_bounded": 0.020941162109375, "reward/mean_retrieval_combined_raw": 0.028907531499862672, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.975, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.975, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1112.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3091322660446167, "rewards/retrieval_reward/mean": 0.9953125, "rewards/retrieval_reward/std": 0.308984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7168916702270507, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.25726017355918884, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.544384002685547, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21153601706027986, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1115, "step_time": 13.750735022011213 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.64921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.2, "completions/min_terminated_length": 0.0, "entropy": 0.4534087084233761, "epoch": 4.426877470355731, "frac_reward_zero_std": 0.0, "grad_norm": 1.9820717573165894, "learning_rate": 2e-05, "loss": 0.000360107421875, "num_tokens": 66860807.0, "reward": 1.011834716796875, "reward/completion_length_max": 36.2, "reward/completion_length_mean": 30.65, "reward/completion_length_min": 27.2, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.00390625, "reward/malformed_ranked_plain_query_rate": 0.00390625, "reward/mean": 1.01171875, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.890625, "reward/mean_r_ndcg": 0.8, "reward/mean_r_ndcg_cutoff": 6.05625, "reward/mean_recall": 0.2453125, "reward/mean_recall_at_10": 0.04892578125, "reward/mean_recall_at_100": 0.2453125, "reward/mean_retrieval_combined_bounded": 0.0226287841796875, "reward/mean_retrieval_combined_raw": 0.0294447124004364, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1118.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3109579384326935, "rewards/retrieval_reward/mean": 1.01171875, "rewards/retrieval_reward/std": 0.311328125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7732115268707274, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21429676711559295, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.341975212097168, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22397168278694152, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1120, "step_time": 13.800192417600192 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 55.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.45859375, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.8, "completions/min_terminated_length": 0.0, "entropy": 0.435075081884861, "epoch": 4.446640316205533, "frac_reward_zero_std": 0.0, "grad_norm": 1.5430030822753906, "learning_rate": 2e-05, "loss": 0.0154296875, "num_tokens": 67157298.0, "reward": 1.0308624267578126, "reward/completion_length_max": 55.0, "reward/completion_length_mean": 30.425, "reward/completion_length_min": 26.8, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.0109375, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.9640625, "reward/mean_mrr": 0.9890625, "reward/mean_ndcg_at_10": 0.8546875, "reward/mean_r_ndcg": 0.73515625, "reward/mean_r_ndcg_cutoff": 5.146875, "reward/mean_recall": 0.2310546875, "reward/mean_recall_at_10": 0.048046875, "reward/mean_recall_at_100": 0.2310546875, "reward/mean_retrieval_combined_bounded": 0.0444580078125, "reward/mean_retrieval_combined_raw": 0.100927734375, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.9890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1120.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.3023688614368439, "rewards/retrieval_reward/mean": 1.0296875, "rewards/retrieval_reward/std": 0.302734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7619665622711183, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.22870997786521913, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.7138420104980465, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21408962905406953, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1125, "step_time": 14.589144374418538 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.2015625, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.2, "completions/min_terminated_length": 0.0, "entropy": 0.4460327818989754, "epoch": 4.466403162055336, "frac_reward_zero_std": 0.0, "grad_norm": 1.8610892295837402, "learning_rate": 2e-05, "loss": -0.000177001953125, "num_tokens": 67453652.0, "reward": 1.004034423828125, "reward/completion_length_max": 35.2, "reward/completion_length_mean": 30.2, "reward/completion_length_min": 27.2, "reward/completion_length_p90": 32.4, "reward/completion_length_p95": 32.8, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00234375, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.89453125, "reward/mean_r_ndcg": 0.75, "reward/mean_r_ndcg_cutoff": 5.51875, "reward/mean_recall": 0.221484375, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.221484375, "reward/mean_retrieval_combined_bounded": 0.0055328369140625, "reward/mean_retrieval_combined_raw": -0.062622781842947, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1128.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2761383533477783, "rewards/retrieval_reward/mean": 1.00234375, "rewards/retrieval_reward/std": 0.2765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.629384088516235, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.22228190004825593, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.689883041381836, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21826221346855162, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1130, "step_time": 13.683636119554285 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.153125, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.4, "completions/min_terminated_length": 0.0, "entropy": 0.40562144592404364, "epoch": 4.486166007905139, "frac_reward_zero_std": 0.0, "grad_norm": 1.5329532623291016, "learning_rate": 2e-05, "loss": -9.65118408203125e-05, "num_tokens": 67749912.0, "reward": 1.0180023193359375, "reward/completion_length_max": 35.2, "reward/completion_length_mean": 30.15, "reward/completion_length_min": 26.4, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.99765625, "reward/mean_ndcg_at_10": 0.88671875, "reward/mean_r_ndcg": 0.77265625, "reward/mean_r_ndcg_cutoff": 5.9375, "reward/mean_recall": 0.2095703125, "reward/mean_recall_at_10": 0.0484375, "reward/mean_recall_at_100": 0.2095703125, "reward/mean_retrieval_combined_bounded": 0.02022705078125, "reward/mean_retrieval_combined_raw": 0.08806232213974, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1132.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2934233546257019, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.29375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8999266147613527, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2537159979343414, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.098996543884278, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2232722520828247, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1135, "step_time": 13.757223559136037 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.66328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.35813857838511465, "epoch": 4.5059288537549405, "frac_reward_zero_std": 0.0, "grad_norm": 1.656117558479309, "learning_rate": 2e-05, "loss": -0.0009033203125, "num_tokens": 68046697.0, "reward": 1.015460205078125, "reward/completion_length_max": 35.4, "reward/completion_length_mean": 30.675, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.975, "reward/mean_mrr": 0.99765625, "reward/mean_ndcg_at_10": 0.9, "reward/mean_r_ndcg": 0.80625, "reward/mean_r_ndcg_cutoff": 6.5125, "reward/mean_recall": 0.245703125, "reward/mean_recall_at_10": 0.049072265625, "reward/mean_recall_at_100": 0.245703125, "reward/mean_retrieval_combined_bounded": 0.0169464111328125, "reward/mean_retrieval_combined_raw": 0.012400132045149804, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1136.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2878503382205963, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.28828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5343973636627197, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2399473637342453, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.37144718170166, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2007121056318283, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1140, "step_time": 13.729951989208349 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.30234375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.38433651998639107, "epoch": 4.525691699604743, "frac_reward_zero_std": 0.0, "grad_norm": 1.9520761966705322, "learning_rate": 2e-05, "loss": -0.00021514892578125, "num_tokens": 68344396.0, "reward": 1.0317413330078125, "reward/completion_length_max": 37.4, "reward/completion_length_mean": 31.3, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.00859375, "reward/duplicate_ranked_plain_query_rate": 0.00859375, "reward/invalid_output_rate": 0.00859375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.96640625, "reward/mean_mrr": 0.99140625, "reward/mean_ndcg_at_10": 0.9109375, "reward/mean_r_ndcg": 0.7671875, "reward/mean_r_ndcg_cutoff": 5.93125, "reward/mean_recall": 0.23125, "reward/mean_recall_at_10": 0.048486328125, "reward/mean_recall_at_100": 0.23125, "reward/mean_retrieval_combined_bounded": 0.0404388427734375, "reward/mean_retrieval_combined_raw": 0.10645285248756409, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99140625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99140625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1142.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2907447159290314, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.29140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.9386165142059326, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21170172691345215, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.435214424133301, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2495229572057724, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1145, "step_time": 13.838200113992206 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.02890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.4, "completions/min_terminated_length": 0.0, "entropy": 0.3650574117898941, "epoch": 4.545454545454545, "frac_reward_zero_std": 0.0, "grad_norm": 1.864538550376892, "learning_rate": 2e-05, "loss": 0.0007415771484375, "num_tokens": 68641457.0, "reward": 1.023260498046875, "reward/completion_length_max": 35.6, "reward/completion_length_mean": 31.05, "reward/completion_length_min": 27.4, "reward/completion_length_p90": 33.4, "reward/completion_length_p95": 34.2, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.01015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.9625, "reward/mean_mrr": 0.98984375, "reward/mean_ndcg_at_10": 0.9046875, "reward/mean_r_ndcg": 0.75234375, "reward/mean_r_ndcg_cutoff": 6.075, "reward/mean_recall": 0.2220703125, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.2220703125, "reward/mean_retrieval_combined_bounded": 0.0333251953125, "reward/mean_retrieval_combined_raw": 0.0524658203125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1147.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2766249179840088, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.2771484375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8060960292816164, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.24675958156585692, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.850078773498534, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21732852458953858, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1150, "step_time": 13.727676730777603 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.23828125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.38967856392264366, "epoch": 4.549407114624506, "frac_reward_zero_std": 0.0, "num_tokens": 68701006.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6658044186731252, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.717969055378479, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8971880492091389, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9956063268892794, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6170314655047207, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.663297791378901, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8266666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.34012808269856715, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.37720151042006744, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5397590361445783, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8530120481927711, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.622322489028694, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6303997607419675, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.754683403451009, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9113640174379611, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.583529743702904, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.36814196466926796, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.18046442975984464, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.32502910809918706, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 2.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3450.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.41696408912563365, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4559706889222166, "retrieval_eval/nq/single_turn/recall_at_10": 0.6598309178743961, "retrieval_eval/nq/single_turn/recall_at_100": 0.8935748792270531, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6704063674859747, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7002047351890938, "retrieval_eval/scifact/single_turn/recall_at_10": 0.834, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9586666666666668, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.03289794921875, "reward/completion_length_max": 36.0, "reward/completion_length_mean": 31.25, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.01953125, "reward/duplicate_ranked_plain_query_rate": 0.01953125, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.95703125, "reward/mean_mrr": 0.98046875, "reward/mean_ndcg_at_10": 0.80078125, "reward/mean_r_ndcg": 0.70703125, "reward/mean_r_ndcg_cutoff": 6.15625, "reward/mean_recall": 0.2373046875, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.2373046875, "reward/mean_retrieval_combined_bounded": 0.052490234375, "reward/mean_retrieval_combined_raw": 0.091796875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1152.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.300220251083374, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.30078125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.527202606201172, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.19385918974876404, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.019094467163086, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25055035948753357, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1151, "step_time": 13.767657254997175 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.75, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.416015625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.25, "completions/min_terminated_length": 0.0, "entropy": 0.39243542309850454, "epoch": 4.565217391304348, "frac_reward_zero_std": 0.0, "grad_norm": 1.6488934755325317, "learning_rate": 2e-05, "loss": -0.0025634765625, "num_tokens": 68939128.0, "reward": 1.0243453979492188, "reward/completion_length_max": 36.75, "reward/completion_length_mean": 31.4375, "reward/completion_length_min": 28.25, "reward/completion_length_p90": 34.25, "reward/completion_length_p95": 34.25, "reward/duplicate_query_rate": 0.013671875, "reward/duplicate_ranked_plain_query_rate": 0.013671875, "reward/invalid_output_rate": 0.013671875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.9609375, "reward/mean_mrr": 0.986328125, "reward/mean_ndcg_at_10": 0.845703125, "reward/mean_r_ndcg": 0.677734375, "reward/mean_r_ndcg_cutoff": 5.9453125, "reward/mean_recall": 0.2412109375, "reward/mean_recall_at_10": 0.04840087890625, "reward/mean_recall_at_100": 0.2412109375, "reward/mean_retrieval_combined_bounded": 0.038043975830078125, "reward/mean_retrieval_combined_raw": 0.05767810670658946, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.986328125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.986328125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1152.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.30813173949718475, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.30810546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7276697754859924, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.22713173180818558, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.08297610282898, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24406441673636436, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1155, "step_time": 13.837296250509098 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.3421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.39115430787205696, "epoch": 4.58498023715415, "frac_reward_zero_std": 0.0, "grad_norm": 1.7532719373703003, "learning_rate": 2e-05, "loss": -0.000726318359375, "num_tokens": 69236974.0, "reward": 1.00499267578125, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 31.325, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.003125, "reward/low_length_query_rate": 0.003125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00546875, "reward/mean_colbert": 0.9671875, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.86484375, "reward/mean_r_ndcg": 0.71484375, "reward/mean_r_ndcg_cutoff": 5.3625, "reward/mean_recall": 0.2181640625, "reward/mean_recall_at_10": 0.0484375, "reward/mean_recall_at_100": 0.2181640625, "reward/mean_retrieval_combined_bounded": 0.0128173828125, "reward/mean_retrieval_combined_raw": 0.028001070022583008, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1156.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.003125, "reward/too_short_query_rate": 0.0, "reward_std": 0.29244797825813296, "rewards/retrieval_reward/mean": 1.00546875, "rewards/retrieval_reward/std": 0.292578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5674150228500365, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1932393103837967, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.74937572479248, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23586967289447786, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1160, "step_time": 13.916233871434816 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.13984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.36153122410178185, "epoch": 4.604743083003952, "frac_reward_zero_std": 0.0, "grad_norm": 2.008509635925293, "learning_rate": 2e-05, "loss": 0.00098876953125, "num_tokens": 69536129.0, "reward": 1.01107177734375, "reward/completion_length_max": 38.2, "reward/completion_length_mean": 32.125, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0109375, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.90078125, "reward/mean_r_ndcg": 0.7875, "reward/mean_r_ndcg_cutoff": 7.06875, "reward/mean_recall": 0.2587890625, "reward/mean_recall_at_10": 0.04912109375, "reward/mean_recall_at_100": 0.2587890625, "reward/mean_retrieval_combined_bounded": 0.01610870361328125, "reward/mean_retrieval_combined_raw": 0.023338764905929565, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1161.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.2866778790950775, "rewards/retrieval_reward/mean": 1.0109375, "rewards/retrieval_reward/std": 0.287109375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5469174861907957, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18880058825016022, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.664632034301757, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24755480587482454, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1165, "step_time": 13.981716727744788 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.1453125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.411263570189476, "epoch": 4.624505928853755, "frac_reward_zero_std": 0.0, "grad_norm": 1.950817346572876, "learning_rate": 2e-05, "loss": 0.0007415771484375, "num_tokens": 69834331.0, "reward": 1.0254913330078126, "reward/completion_length_max": 37.2, "reward/completion_length_mean": 32.1, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.0125, "reward/duplicate_ranked_plain_query_rate": 0.0125, "reward/invalid_output_rate": 0.0125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.9875, "reward/mean_ndcg_at_10": 0.875, "reward/mean_r_ndcg": 0.77265625, "reward/mean_r_ndcg_cutoff": 6.675, "reward/mean_recall": 0.2474609375, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.2474609375, "reward/mean_retrieval_combined_bounded": 0.038079833984375, "reward/mean_retrieval_combined_raw": 0.054052571207284926, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1168.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2875346541404724, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.2875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7342761516571046, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20506676733493806, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.472177696228027, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24097910821437835, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1170, "step_time": 13.812281670840457 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.8890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.8, "completions/min_terminated_length": 0.0, "entropy": 0.3972558543086052, "epoch": 4.644268774703558, "frac_reward_zero_std": 0.0, "grad_norm": 1.9129767417907715, "learning_rate": 2e-05, "loss": 0.0004852294921875, "num_tokens": 70134093.0, "reward": 1.02421875, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 32.9, "reward/completion_length_min": 29.8, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.9296875, "reward/mean_r_ndcg": 0.76484375, "reward/mean_r_ndcg_cutoff": 7.5625, "reward/mean_recall": 0.2361328125, "reward/mean_recall_at_10": 0.049609375, "reward/mean_recall_at_100": 0.2361328125, "reward/mean_retrieval_combined_bounded": 0.0285400390625, "reward/mean_retrieval_combined_raw": 0.016255798935890197, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1171.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.2924102544784546, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.292578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6290019512176515, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.17161630392074584, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.327583312988281, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27192283272743223, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1175, "step_time": 13.887983610422816 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.784375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.4171502165496349, "epoch": 4.664031620553359, "frac_reward_zero_std": 0.0, "grad_norm": 1.8693459033966064, "learning_rate": 2e-05, "loss": 0.00013427734375, "num_tokens": 70433817.0, "reward": 1.0263946533203125, "reward/completion_length_max": 37.4, "reward/completion_length_mean": 32.825, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.003125, "reward/low_length_query_rate": 0.003125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.96640625, "reward/mean_mrr": 0.990625, "reward/mean_ndcg_at_10": 0.8859375, "reward/mean_r_ndcg": 0.78203125, "reward/mean_r_ndcg_cutoff": 6.28125, "reward/mean_recall": 0.237109375, "reward/mean_recall_at_10": 0.048291015625, "reward/mean_recall_at_100": 0.237109375, "reward/mean_retrieval_combined_bounded": 0.03720703125, "reward/mean_retrieval_combined_raw": 0.04550778325647116, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1176.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.003125, "reward/too_short_query_rate": 0.0, "reward_std": 0.28568325042724607, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.2859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.519251489639282, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1801351249217987, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.1120210647583, "sampling/raw_vllm_sampling_logp_difference/mean": 0.257817742228508, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1180, "step_time": 13.838843207398895 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.43203125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.4189584098756313, "epoch": 4.683794466403162, "frac_reward_zero_std": 0.0, "grad_norm": 2.0024421215057373, "learning_rate": 2e-05, "loss": 0.0001708984375, "num_tokens": 70732930.0, "reward": 1.024261474609375, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 32.425, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 34.8, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00859375, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99140625, "reward/mean_ndcg_at_10": 0.9265625, "reward/mean_r_ndcg": 0.8671875, "reward/mean_r_ndcg_cutoff": 7.275, "reward/mean_recall": 0.2369140625, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.2369140625, "reward/mean_retrieval_combined_bounded": 0.03388671875, "reward/mean_retrieval_combined_raw": 0.03794074058532715, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99140625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99140625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1184.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.2827356159687042, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.283203125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.521680164337158, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15867041647434235, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.302172470092774, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2677730917930603, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1185, "step_time": 13.918280883401167 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.4515625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.8, "completions/min_terminated_length": 0.0, "entropy": 0.39549348428845404, "epoch": 4.703557312252965, "frac_reward_zero_std": 0.0, "grad_norm": 1.766048789024353, "learning_rate": 2e-05, "loss": 0.0008544921875, "num_tokens": 71031812.0, "reward": 0.9871490478515625, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 32.5, "reward/completion_length_min": 28.8, "reward/completion_length_p90": 36.2, "reward/completion_length_p95": 36.8, "reward/duplicate_query_rate": 0.0171875, "reward/duplicate_ranked_plain_query_rate": 0.0171875, "reward/invalid_output_rate": 0.03984375, "reward/length_out_of_range_rate": 0.02265625, "reward/low_length_query_rate": 0.02265625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.9875, "reward/mean_colbert": 0.9375, "reward/mean_mrr": 0.96015625, "reward/mean_ndcg_at_10": 0.86328125, "reward/mean_r_ndcg": 0.7640625, "reward/mean_r_ndcg_cutoff": 5.41875, "reward/mean_recall": 0.21953125, "reward/mean_recall_at_10": 0.046826171875, "reward/mean_recall_at_100": 0.21953125, "reward/mean_retrieval_combined_bounded": 0.0382568359375, "reward/mean_retrieval_combined_raw": 0.07861328125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.96015625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96015625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1185.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.02265625, "reward/too_short_query_rate": 0.0, "reward_std": 0.33885252475738525, "rewards/retrieval_reward/mean": 0.9875, "rewards/retrieval_reward/std": 0.3390625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.771527910232544, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16863133609294892, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.298305892944336, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2757857084274292, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1190, "step_time": 13.887304081604816 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.2453125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.8, "completions/min_terminated_length": 0.0, "entropy": 0.38728041872382163, "epoch": 4.723320158102767, "frac_reward_zero_std": 0.0, "grad_norm": 2.0961694717407227, "learning_rate": 2e-05, "loss": 0.001708984375, "num_tokens": 71330590.0, "reward": 1.0204681396484374, "reward/completion_length_max": 37.2, "reward/completion_length_mean": 32.275, "reward/completion_length_min": 28.8, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.01171875, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.00390625, "reward/malformed_ranked_plain_query_rate": 0.00390625, "reward/mean": 1.0203125, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.903125, "reward/mean_r_ndcg": 0.7796875, "reward/mean_r_ndcg_cutoff": 6.16875, "reward/mean_recall": 0.2470703125, "reward/mean_recall_at_10": 0.048974609375, "reward/mean_recall_at_100": 0.2470703125, "reward/mean_retrieval_combined_bounded": 0.03656005859375, "reward/mean_retrieval_combined_raw": 0.05132176280021668, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1192.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.31217005252838137, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.31328125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.794326829910278, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.14418634474277497, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.785591888427734, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26702251434326174, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1195, "step_time": 13.817429067369085 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.58125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.37591483891010286, "epoch": 4.743083003952569, "frac_reward_zero_std": 0.0, "grad_norm": 2.441627264022827, "learning_rate": 2e-05, "loss": -0.001031494140625, "num_tokens": 71628742.0, "reward": 1.0151275634765624, "reward/completion_length_max": 36.4, "reward/completion_length_mean": 31.575, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 33.6, "reward/completion_length_p95": 34.6, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.97421875, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.9140625, "reward/mean_r_ndcg": 0.7859375, "reward/mean_r_ndcg_cutoff": 6.8625, "reward/mean_recall": 0.2251953125, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.2251953125, "reward/mean_retrieval_combined_bounded": 0.015997314453125, "reward/mean_retrieval_combined_raw": 0.006348994188010692, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1198.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2841807723045349, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.284765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8125601768493653, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.19869315922260283, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.909850883483887, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2376388728618622, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1200, "step_time": 13.835678840172477 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 43.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.1796875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.3507183752954006, "epoch": 4.747035573122529, "frac_reward_zero_std": 0.0, "num_tokens": 71688724.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.677218262711837, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7299603599772978, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9086115992970123, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9947275922671354, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6180011572645782, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6565418635721605, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3348329911556989, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.37363370501578214, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5397590361445783, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8554216867469879, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6339202791982526, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6416756670974345, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7647018474483264, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9074237089201878, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5666826133879168, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3577963708290441, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17427407675250314, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3146019918208786, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 1.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3451.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.41725853156060866, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4557975604142653, "retrieval_eval/nq/single_turn/recall_at_10": 0.6613300492610837, "retrieval_eval/nq/single_turn/recall_at_100": 0.8929778808074954, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6747293108761127, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7072481398121591, "retrieval_eval/scifact/single_turn/recall_at_10": 0.844, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9470000000000001, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 0.99481201171875, "reward/completion_length_max": 43.0, "reward/completion_length_mean": 33.25, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 38.0, "reward/completion_length_p95": 40.0, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.046875, "reward/length_out_of_range_rate": 0.0390625, "reward/low_length_query_rate": 0.0390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99609375, "reward/mean_colbert": 0.9296875, "reward/mean_mrr": 0.953125, "reward/mean_ndcg_at_10": 0.8984375, "reward/mean_r_ndcg": 0.78125, "reward/mean_r_ndcg_cutoff": 6.0, "reward/mean_recall": 0.197265625, "reward/mean_recall_at_10": 0.047607421875, "reward/mean_recall_at_100": 0.197265625, "reward/mean_retrieval_combined_bounded": 0.061279296875, "reward/mean_retrieval_combined_raw": 0.12890625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1200.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0390625, "reward/too_short_query_rate": 0.0, "reward_std": 0.4227191209793091, "rewards/retrieval_reward/mean": 0.99609375, "rewards/retrieval_reward/std": 0.423828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3343381881713867, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16425254940986633, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.654993057250977, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27874094247817993, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1201, "step_time": 14.068766948068514 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.587890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.75, "completions/min_terminated_length": 0.0, "entropy": 0.3944906983524561, "epoch": 4.762845849802371, "frac_reward_zero_std": 0.0, "grad_norm": 2.088275671005249, "learning_rate": 2e-05, "loss": 0.00341796875, "num_tokens": 71926862.0, "reward": 1.0051536560058594, "reward/completion_length_max": 39.0, "reward/completion_length_mean": 31.59375, "reward/completion_length_min": 27.75, "reward/completion_length_p90": 34.75, "reward/completion_length_p95": 35.5, "reward/duplicate_query_rate": 0.0361328125, "reward/duplicate_ranked_plain_query_rate": 0.0361328125, "reward/invalid_output_rate": 0.0390625, "reward/length_out_of_range_rate": 0.0029296875, "reward/low_length_query_rate": 0.0029296875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.005859375, "reward/mean_colbert": 0.939453125, "reward/mean_mrr": 0.9609375, "reward/mean_ndcg_at_10": 0.8720703125, "reward/mean_r_ndcg": 0.7724609375, "reward/mean_r_ndcg_cutoff": 5.8515625, "reward/mean_recall": 0.229248046875, "reward/mean_recall_at_10": 0.04705810546875, "reward/mean_recall_at_100": 0.229248046875, "reward/mean_retrieval_combined_bounded": 0.04555511474609375, "reward/mean_retrieval_combined_raw": 0.08889460563659668, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1200.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0029296875, "reward/too_short_query_rate": 0.0, "reward_std": 0.3232552632689476, "rewards/retrieval_reward/mean": 1.005859375, "rewards/retrieval_reward/std": 0.32373046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.2389723658561707, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.189080648124218, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.44884967803955, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24496164917945862, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1205, "step_time": 13.85558746921015 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.89296875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.4, "completions/min_terminated_length": 0.0, "entropy": 0.37455667853355407, "epoch": 4.782608695652174, "frac_reward_zero_std": 0.0, "grad_norm": 1.9337142705917358, "learning_rate": 2e-05, "loss": -0.0004364013671875, "num_tokens": 72223909.0, "reward": 1.0118408203125, "reward/completion_length_max": 35.0, "reward/completion_length_mean": 30.875, "reward/completion_length_min": 27.4, "reward/completion_length_p90": 32.8, "reward/completion_length_p95": 33.4, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0125, "reward/mean_colbert": 0.9703125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.9109375, "reward/mean_r_ndcg": 0.74921875, "reward/mean_r_ndcg_cutoff": 5.85625, "reward/mean_recall": 0.2404296875, "reward/mean_recall_at_10": 0.0490234375, "reward/mean_recall_at_100": 0.2404296875, "reward/mean_retrieval_combined_bounded": 0.018109893798828124, "reward/mean_retrieval_combined_raw": 0.016022347658872605, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1208.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2811030626296997, "rewards/retrieval_reward/mean": 1.0125, "rewards/retrieval_reward/std": 0.28125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8570135593414308, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2341648519039154, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.823244380950928, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23085922002792358, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1210, "step_time": 13.717297561047598 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.2625, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.2, "completions/min_terminated_length": 0.0, "entropy": 0.38385465145111086, "epoch": 4.8023715415019765, "frac_reward_zero_std": 0.0, "grad_norm": 1.9245156049728394, "learning_rate": 2e-05, "loss": 0.001116943359375, "num_tokens": 72520245.0, "reward": 1.009820556640625, "reward/completion_length_max": 35.4, "reward/completion_length_mean": 30.275, "reward/completion_length_min": 27.2, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 32.8, "reward/duplicate_query_rate": 0.009375, "reward/duplicate_ranked_plain_query_rate": 0.009375, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01015625, "reward/mean_colbert": 0.96640625, "reward/mean_mrr": 0.990625, "reward/mean_ndcg_at_10": 0.89921875, "reward/mean_r_ndcg": 0.8234375, "reward/mean_r_ndcg_cutoff": 6.59375, "reward/mean_recall": 0.2341796875, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.2341796875, "reward/mean_retrieval_combined_bounded": 0.01931915283203125, "reward/mean_retrieval_combined_raw": 0.03311706557869911, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1212.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2889146327972412, "rewards/retrieval_reward/mean": 1.01015625, "rewards/retrieval_reward/std": 0.289453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4815134048461913, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18629871308803558, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.498547554016113, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23012405037879943, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1215, "step_time": 13.71943800660083 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.99921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.0, "completions/min_terminated_length": 0.0, "entropy": 0.40586271733045576, "epoch": 4.822134387351778, "frac_reward_zero_std": 0.0, "grad_norm": 2.7398476600646973, "learning_rate": 2e-05, "loss": -0.0017578125, "num_tokens": 72815860.0, "reward": 1.0216522216796875, "reward/completion_length_max": 36.4, "reward/completion_length_mean": 29.975, "reward/completion_length_min": 26.0, "reward/completion_length_p90": 33.6, "reward/completion_length_p95": 34.2, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.87734375, "reward/mean_r_ndcg": 0.78515625, "reward/mean_r_ndcg_cutoff": 6.2, "reward/mean_recall": 0.230078125, "reward/mean_recall_at_10": 0.047412109375, "reward/mean_recall_at_100": 0.230078125, "reward/mean_retrieval_combined_bounded": 0.0255126953125, "reward/mean_retrieval_combined_raw": 0.02778577208518982, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1216.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2787415862083435, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.27890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8135658740997314, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.22012647092342377, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.09115161895752, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23562210202217101, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1220, "step_time": 13.807261043635663 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.04765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.6, "completions/min_terminated_length": 0.0, "entropy": 0.3764793246984482, "epoch": 4.841897233201581, "frac_reward_zero_std": 0.0, "grad_norm": 2.0205483436584473, "learning_rate": 2e-05, "loss": -0.001324462890625, "num_tokens": 73111665.0, "reward": 1.0089752197265625, "reward/completion_length_max": 34.8, "reward/completion_length_mean": 30.075, "reward/completion_length_min": 26.6, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.009375, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.90078125, "reward/mean_r_ndcg": 0.7171875, "reward/mean_r_ndcg_cutoff": 6.15, "reward/mean_recall": 0.2181640625, "reward/mean_recall_at_10": 0.048046875, "reward/mean_recall_at_100": 0.2181640625, "reward/mean_retrieval_combined_bounded": 0.0128509521484375, "reward/mean_retrieval_combined_raw": 0.027002468705177307, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1222.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2786727428436279, "rewards/retrieval_reward/mean": 1.009375, "rewards/retrieval_reward/std": 0.27890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.759205675125122, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.17794448882341385, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.676036071777344, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2582526713609695, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1225, "step_time": 13.753363653214183 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.74140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.3736093960702419, "epoch": 4.861660079051384, "frac_reward_zero_std": 0.0, "grad_norm": 1.8976631164550781, "learning_rate": 2e-05, "loss": 0.002099609375, "num_tokens": 73408454.0, "reward": 1.0238006591796875, "reward/completion_length_max": 35.8, "reward/completion_length_mean": 30.725, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.6, "reward/duplicate_query_rate": 0.01640625, "reward/duplicate_ranked_plain_query_rate": 0.01640625, "reward/invalid_output_rate": 0.01640625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.959375, "reward/mean_mrr": 0.9828125, "reward/mean_ndcg_at_10": 0.884375, "reward/mean_r_ndcg": 0.77578125, "reward/mean_r_ndcg_cutoff": 5.8875, "reward/mean_recall": 0.232421875, "reward/mean_recall_at_10": 0.0478515625, "reward/mean_recall_at_100": 0.232421875, "reward/mean_retrieval_combined_bounded": 0.0403076171875, "reward/mean_retrieval_combined_raw": 0.07089895308017731, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98359375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98359375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1227.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2968686640262604, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.297265625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6323357582092286, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16569003015756606, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.403483009338379, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27160250544548037, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1230, "step_time": 13.72124032441061 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.60078125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.42282912880182266, "epoch": 4.881422924901186, "frac_reward_zero_std": 0.0, "grad_norm": 1.9893394708633423, "learning_rate": 2e-05, "loss": -0.000274658203125, "num_tokens": 73706535.0, "reward": 1.0067169189453125, "reward/completion_length_max": 37.6, "reward/completion_length_mean": 31.625, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00625, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.89765625, "reward/mean_r_ndcg": 0.79140625, "reward/mean_r_ndcg_cutoff": 6.65, "reward/mean_recall": 0.2572265625, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.2572265625, "reward/mean_retrieval_combined_bounded": 0.010595703125, "reward/mean_retrieval_combined_raw": -0.055053113400936125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1232.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.28878192901611327, "rewards/retrieval_reward/mean": 1.00625, "rewards/retrieval_reward/std": 0.289453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5574661016464235, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1739799588918686, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.166439056396484, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26091599762439727, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1235, "step_time": 13.87942900257185 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.884375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.4, "completions/min_terminated_length": 0.0, "entropy": 0.43915337696671486, "epoch": 4.901185770750988, "frac_reward_zero_std": 0.0, "grad_norm": 1.9582067728042603, "learning_rate": 2e-05, "loss": 0.00026702880859375, "num_tokens": 74005715.0, "reward": 1.0195068359375, "reward/completion_length_max": 37.6, "reward/completion_length_mean": 32.85, "reward/completion_length_min": 29.4, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.0265625, "reward/duplicate_ranked_plain_query_rate": 0.0265625, "reward/invalid_output_rate": 0.0265625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.94921875, "reward/mean_mrr": 0.9734375, "reward/mean_ndcg_at_10": 0.86484375, "reward/mean_r_ndcg": 0.75234375, "reward/mean_r_ndcg_cutoff": 6.21875, "reward/mean_recall": 0.2369140625, "reward/mean_recall_at_10": 0.0478515625, "reward/mean_recall_at_100": 0.2369140625, "reward/mean_retrieval_combined_bounded": 0.04608154296875, "reward/mean_retrieval_combined_raw": 0.08469545841217041, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9734375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9734375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1236.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3143456280231476, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.315234375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.786379337310791, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15336838066577912, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.50786075592041, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2751438468694687, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1240, "step_time": 13.817102053610142 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.2390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.6, "completions/min_terminated_length": 0.0, "entropy": 0.4322821117937565, "epoch": 4.920948616600791, "frac_reward_zero_std": 0.0, "grad_norm": 2.175813913345337, "learning_rate": 2e-05, "loss": 0.001318359375, "num_tokens": 74307141.0, "reward": 0.9896575927734375, "reward/completion_length_max": 40.2, "reward/completion_length_mean": 34.3, "reward/completion_length_min": 30.6, "reward/completion_length_p90": 36.8, "reward/completion_length_p95": 37.4, "reward/duplicate_query_rate": 0.02265625, "reward/duplicate_ranked_plain_query_rate": 0.02265625, "reward/invalid_output_rate": 0.0328125, "reward/length_out_of_range_rate": 0.01015625, "reward/low_length_query_rate": 0.01015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.9890625, "reward/mean_colbert": 0.9421875, "reward/mean_mrr": 0.9671875, "reward/mean_ndcg_at_10": 0.8625, "reward/mean_r_ndcg": 0.619140625, "reward/mean_r_ndcg_cutoff": 5.00625, "reward/mean_recall": 0.2177734375, "reward/mean_recall_at_10": 0.047265625, "reward/mean_recall_at_100": 0.2177734375, "reward/mean_retrieval_combined_bounded": 0.027508544921875, "reward/mean_retrieval_combined_raw": 0.07002018899656832, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9671875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9671875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1241.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.33004379868507383, "rewards/retrieval_reward/mean": 0.990625, "rewards/retrieval_reward/std": 0.330078125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.656083345413208, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12865542024374008, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.962148475646973, "sampling/raw_vllm_sampling_logp_difference/mean": 0.28836464285850527, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1245, "step_time": 13.974556808487979 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.57578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.6, "completions/min_terminated_length": 0.0, "entropy": 0.42548945769667623, "epoch": 4.940711462450593, "frac_reward_zero_std": 0.0, "grad_norm": 1.821857213973999, "learning_rate": 2e-05, "loss": 0.0006622314453125, "num_tokens": 74607398.0, "reward": 1.02099609375, "reward/completion_length_max": 39.2, "reward/completion_length_mean": 33.6, "reward/completion_length_min": 29.6, "reward/completion_length_p90": 36.2, "reward/completion_length_p95": 36.8, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.00859375, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.96640625, "reward/mean_mrr": 0.99140625, "reward/mean_ndcg_at_10": 0.8984375, "reward/mean_r_ndcg": 0.7359375, "reward/mean_r_ndcg_cutoff": 5.70625, "reward/mean_recall": 0.2193359375, "reward/mean_recall_at_10": 0.04873046875, "reward/mean_recall_at_100": 0.2193359375, "reward/mean_retrieval_combined_bounded": 0.0302001953125, "reward/mean_retrieval_combined_raw": 0.0572265625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99140625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99140625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1248.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.2874920070171356, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.2880859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.267549753189087, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12621051520109178, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.845100593566894, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2923040747642517, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1250, "step_time": 13.986956582008862 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.34765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.0, "completions/min_terminated_length": 0.0, "entropy": 0.4228396862745285, "epoch": 4.944664031620554, "frac_reward_zero_std": 0.0, "num_tokens": 74667359.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6656005861042831, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7173450722011788, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8954305799648506, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9947275922671354, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5988802618649874, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.656039736427466, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8533333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.33124616705108423, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3647224247596083, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5156626506024097, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8530120481927711, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6281296567180007, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6401843025844954, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7802435827083715, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9083962105969148, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 1.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 322.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5811344741783973, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.36570055525263206, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17994238487786746, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.30948641822087, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 9.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3443.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.41700135915945047, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.45371987217728954, "retrieval_eval/nq/single_turn/recall_at_10": 0.6533304288895343, "retrieval_eval/nq/single_turn/recall_at_100": 0.8868477103301384, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6664110209340213, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7012219080127515, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8383333333333334, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9533333333333334, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0448455810546875, "reward/completion_length_max": 37.0, "reward/completion_length_mean": 33.25, "reward/completion_length_min": 30.0, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.046875, "reward/mean_colbert": 0.9765625, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.87109375, "reward/mean_r_ndcg": 0.7578125, "reward/mean_r_ndcg_cutoff": 7.375, "reward/mean_recall": 0.2451171875, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.2451171875, "reward/mean_retrieval_combined_bounded": 0.044677734375, "reward/mean_retrieval_combined_raw": 0.1865234375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1248.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3123939037322998, "rewards/retrieval_reward/mean": 1.046875, "rewards/retrieval_reward/std": 0.3125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.619489908218384, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1788121461868286, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.097731590270996, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24121050536632538, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1251, "step_time": 13.791698960063513 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.20703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.25, "completions/min_terminated_length": 0.0, "entropy": 0.450278053060174, "epoch": 4.9604743083003955, "frac_reward_zero_std": 0.0, "grad_norm": 1.8743479251861572, "learning_rate": 2e-05, "loss": -0.000445556640625, "num_tokens": 74907603.0, "reward": 1.0115127563476562, "reward/completion_length_max": 39.25, "reward/completion_length_mean": 33.25, "reward/completion_length_min": 29.25, "reward/completion_length_p90": 35.75, "reward/completion_length_p95": 36.25, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.001953125, "reward/length_out_of_range_rate": 0.001953125, "reward/low_length_query_rate": 0.001953125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.009765625, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.998046875, "reward/mean_ndcg_at_10": 0.91796875, "reward/mean_r_ndcg": 0.8134765625, "reward/mean_r_ndcg_cutoff": 6.3671875, "reward/mean_recall": 0.237548828125, "reward/mean_recall_at_10": 0.04840087890625, "reward/mean_recall_at_100": 0.237548828125, "reward/mean_retrieval_combined_bounded": 0.0144805908203125, "reward/mean_retrieval_combined_raw": 0.010680438252165914, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.998046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.998046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1252.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.001953125, "reward/too_short_query_rate": 0.0, "reward_std": 0.30457504838705063, "rewards/retrieval_reward/mean": 1.009765625, "rewards/retrieval_reward/std": 0.30517578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7736817598342896, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1818046122789383, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.070854425430298, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2669597528874874, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1255, "step_time": 14.06798620952759 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.559375, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.2, "completions/min_terminated_length": 0.0, "entropy": 0.4093630865216255, "epoch": 4.980237154150197, "frac_reward_zero_std": 0.0, "grad_norm": 2.302952527999878, "learning_rate": 2e-05, "loss": 0.00086669921875, "num_tokens": 75209631.0, "reward": 1.0130096435546876, "reward/completion_length_max": 40.4, "reward/completion_length_mean": 34.55, "reward/completion_length_min": 30.2, "reward/completion_length_p90": 37.2, "reward/completion_length_p95": 38.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.00546875, "reward/low_length_query_rate": 0.00546875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0125, "reward/mean_colbert": 0.965625, "reward/mean_mrr": 0.990625, "reward/mean_ndcg_at_10": 0.86484375, "reward/mean_r_ndcg": 0.675, "reward/mean_r_ndcg_cutoff": 6.24375, "reward/mean_recall": 0.2369140625, "reward/mean_recall_at_10": 0.048876953125, "reward/mean_recall_at_100": 0.2369140625, "reward/mean_retrieval_combined_bounded": 0.02513427734375, "reward/mean_retrieval_combined_raw": 0.0511962890625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1256.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00546875, "reward/too_short_query_rate": 0.0, "reward_std": 0.2963612377643585, "rewards/retrieval_reward/mean": 1.0125, "rewards/retrieval_reward/std": 0.296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7033141613006593, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1667373299598694, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.945645141601563, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25267096757888796, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1260, "step_time": 14.035206253570504 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 35.73984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.8, "completions/min_terminated_length": 0.0, "entropy": 0.40080643743276595, "epoch": 5.0, "frac_reward_zero_std": 0.0, "grad_norm": 1.7897075414657593, "learning_rate": 2e-05, "loss": -9.9945068359375e-05, "num_tokens": 75512882.0, "reward": 1.0077301025390626, "reward/completion_length_max": 40.2, "reward/completion_length_mean": 35.7, "reward/completion_length_min": 31.8, "reward/completion_length_p90": 38.0, "reward/completion_length_p95": 38.6, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.01484375, "reward/length_out_of_range_rate": 0.0078125, "reward/low_length_query_rate": 0.0078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0078125, "reward/mean_colbert": 0.9609375, "reward/mean_mrr": 0.984375, "reward/mean_ndcg_at_10": 0.86328125, "reward/mean_r_ndcg": 0.715625, "reward/mean_r_ndcg_cutoff": 5.8125, "reward/mean_recall": 0.23359375, "reward/mean_recall_at_10": 0.0478515625, "reward/mean_recall_at_100": 0.23359375, "reward/mean_retrieval_combined_bounded": 0.0264892578125, "reward/mean_retrieval_combined_raw": 0.0678955078125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98515625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98515625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1264.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.30861458778381345, "rewards/retrieval_reward/mean": 1.0078125, "rewards/retrieval_reward/std": 0.308984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.669740629196167, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.17898308783769606, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.494124984741211, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23216497898101807, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1265, "step_time": 14.05765495704254 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 43.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.45, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.8, "completions/min_terminated_length": 0.0, "entropy": 0.4065044172108173, "epoch": 5.019762845849803, "frac_reward_zero_std": 0.05, "grad_norm": 1.9810250997543335, "learning_rate": 2e-05, "loss": 0.00625, "num_tokens": 75816946.0, "reward": 0.8645965576171875, "reward/completion_length_max": 43.4, "reward/completion_length_mean": 36.45, "reward/completion_length_min": 31.8, "reward/completion_length_p90": 40.2, "reward/completion_length_p95": 41.2, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.11875, "reward/length_out_of_range_rate": 0.11484375, "reward/low_length_query_rate": 0.11484375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.86484375, "reward/mean_colbert": 0.85859375, "reward/mean_mrr": 0.86953125, "reward/mean_ndcg_at_10": 0.81015625, "reward/mean_r_ndcg": 0.66875, "reward/mean_r_ndcg_cutoff": 5.015625, "reward/mean_recall": 0.1935546875, "reward/mean_recall_at_10": 0.04287109375, "reward/mean_recall_at_100": 0.1935546875, "reward/mean_retrieval_combined_bounded": 0.04078369140625, "reward/mean_retrieval_combined_raw": 0.10429687928408385, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.88125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.88125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1265.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.11484375, "reward/too_short_query_rate": 0.0, "reward_std": 0.4744517683982849, "rewards/retrieval_reward/mean": 0.86484375, "rewards/retrieval_reward/std": 0.475, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7730432510375977, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16947613060474395, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.073800659179687, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24452911019325257, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1270, "step_time": 14.07401546217734 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 42.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.86953125, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.0, "completions/min_terminated_length": 0.0, "entropy": 0.36995776146650317, "epoch": 5.0395256916996045, "frac_reward_zero_std": 0.025, "grad_norm": 1.8779665231704712, "learning_rate": 2e-05, "loss": 0.0034912109375, "num_tokens": 76119179.0, "reward": 0.967816162109375, "reward/completion_length_max": 42.0, "reward/completion_length_mean": 34.95, "reward/completion_length_min": 30.0, "reward/completion_length_p90": 37.6, "reward/completion_length_p95": 38.8, "reward/duplicate_query_rate": 0.01953125, "reward/duplicate_ranked_plain_query_rate": 0.01953125, "reward/invalid_output_rate": 0.06484375, "reward/length_out_of_range_rate": 0.0453125, "reward/low_length_query_rate": 0.0453125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.96796875, "reward/mean_colbert": 0.91171875, "reward/mean_mrr": 0.93515625, "reward/mean_ndcg_at_10": 0.8109375, "reward/mean_r_ndcg": 0.69296875, "reward/mean_r_ndcg_cutoff": 5.45625, "reward/mean_recall": 0.2296875, "reward/mean_recall_at_10": 0.046240234375, "reward/mean_recall_at_100": 0.2296875, "reward/mean_retrieval_combined_bounded": 0.055419921875, "reward/mean_retrieval_combined_raw": 0.1453125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.93515625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.93515625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1272.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0453125, "reward/too_short_query_rate": 0.0, "reward_std": 0.41367809772491454, "rewards/retrieval_reward/mean": 0.96796875, "rewards/retrieval_reward/std": 0.4140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7026399612426757, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21986074149608612, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.868516731262208, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2080828219652176, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1275, "step_time": 13.993628105719107 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.24609375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.33372599110007284, "epoch": 5.059288537549407, "frac_reward_zero_std": 0.0, "grad_norm": 1.719088077545166, "learning_rate": 2e-05, "loss": 0.0041015625, "num_tokens": 76420358.0, "reward": 1.0157073974609374, "reward/completion_length_max": 41.4, "reward/completion_length_mean": 34.25, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 37.4, "reward/completion_length_p95": 38.6, "reward/duplicate_query_rate": 0.01640625, "reward/duplicate_ranked_plain_query_rate": 0.01640625, "reward/invalid_output_rate": 0.025, "reward/length_out_of_range_rate": 0.00859375, "reward/low_length_query_rate": 0.00859375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.9515625, "reward/mean_mrr": 0.975, "reward/mean_ndcg_at_10": 0.89140625, "reward/mean_r_ndcg": 0.81328125, "reward/mean_r_ndcg_cutoff": 6.225, "reward/mean_recall": 0.2375, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.2375, "reward/mean_retrieval_combined_bounded": 0.0448974609375, "reward/mean_retrieval_combined_raw": 0.07744733095169068, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.975, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.975, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1278.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00859375, "reward/too_short_query_rate": 0.0, "reward_std": 0.3209036707878113, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.321875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7903420448303224, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21656689643859864, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.98682622909546, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21589750945568084, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1280, "step_time": 14.053240458399523 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 57.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.40703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.4, "completions/min_terminated_length": 0.0, "entropy": 0.3750967137515545, "epoch": 5.07905138339921, "frac_reward_zero_std": 0.0, "grad_norm": 1.8701856136322021, "learning_rate": 2e-05, "loss": 0.01376953125, "num_tokens": 76720143.0, "reward": 1.0259765625, "reward/completion_length_max": 57.2, "reward/completion_length_mean": 33.4, "reward/completion_length_min": 29.4, "reward/completion_length_p90": 36.4, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.8953125, "reward/mean_r_ndcg": 0.83046875, "reward/mean_r_ndcg_cutoff": 6.4625, "reward/mean_recall": 0.219921875, "reward/mean_recall_at_10": 0.048046875, "reward/mean_recall_at_100": 0.219921875, "reward/mean_retrieval_combined_bounded": 0.03246002197265625, "reward/mean_retrieval_combined_raw": 0.0342296838760376, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1280.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.29009988307952883, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.290234375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7655386924743652, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2126348316669464, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.497112464904785, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21415150463581084, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1285, "step_time": 14.748237343202344 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.0984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.2, "completions/min_terminated_length": 0.0, "entropy": 0.37735470533370974, "epoch": 5.098814229249012, "frac_reward_zero_std": 0.0, "grad_norm": 2.2337958812713623, "learning_rate": 2e-05, "loss": 0.00125732421875, "num_tokens": 77019885.0, "reward": 0.99173583984375, "reward/completion_length_max": 39.6, "reward/completion_length_mean": 33.05, "reward/completion_length_min": 29.2, "reward/completion_length_p90": 36.4, "reward/completion_length_p95": 36.8, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.0296875, "reward/length_out_of_range_rate": 0.021875, "reward/low_length_query_rate": 0.021875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.990625, "reward/mean_colbert": 0.94453125, "reward/mean_mrr": 0.9703125, "reward/mean_ndcg_at_10": 0.89453125, "reward/mean_r_ndcg": 0.74375, "reward/mean_r_ndcg_cutoff": 6.25625, "reward/mean_recall": 0.2208984375, "reward/mean_recall_at_10": 0.047998046875, "reward/mean_recall_at_100": 0.2208984375, "reward/mean_retrieval_combined_bounded": 0.0322509765625, "reward/mean_retrieval_combined_raw": 0.04902656078338623, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9703125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9703125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1288.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.021875, "reward/too_short_query_rate": 0.0, "reward_std": 0.358547705411911, "rewards/retrieval_reward/mean": 0.990625, "rewards/retrieval_reward/std": 0.359765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.900747060775757, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2412216156721115, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.5588476181030275, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21701061725616455, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1290, "step_time": 13.874607194424607 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.11015625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.33621482700109484, "epoch": 5.118577075098814, "frac_reward_zero_std": 0.0, "grad_norm": 1.8332444429397583, "learning_rate": 2e-05, "loss": -0.0005523681640625, "num_tokens": 77318586.0, "reward": 1.009716796875, "reward/completion_length_max": 36.6, "reward/completion_length_mean": 32.125, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 35.4, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.009375, "reward/mean_colbert": 0.97421875, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.88671875, "reward/mean_r_ndcg": 0.721875, "reward/mean_r_ndcg_cutoff": 5.11875, "reward/mean_recall": 0.2203125, "reward/mean_recall_at_10": 0.0490234375, "reward/mean_recall_at_100": 0.2203125, "reward/mean_retrieval_combined_bounded": 0.0105133056640625, "reward/mean_retrieval_combined_raw": 0.006251691654324531, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1292.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2694634974002838, "rewards/retrieval_reward/mean": 1.009375, "rewards/retrieval_reward/std": 0.26953125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6720630168914794, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2103344827890396, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.493767929077148, "sampling/raw_vllm_sampling_logp_difference/mean": 0.224147430062294, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1295, "step_time": 13.879903892625588 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.1484375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.2933156456798315, "epoch": 5.138339920948616, "frac_reward_zero_std": 0.0, "grad_norm": 2.0744361877441406, "learning_rate": 2e-05, "loss": -0.00206298828125, "num_tokens": 77617240.0, "reward": 1.0163238525390625, "reward/completion_length_max": 37.4, "reward/completion_length_mean": 32.15, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 34.8, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.0125, "reward/duplicate_ranked_plain_query_rate": 0.0125, "reward/invalid_output_rate": 0.0125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.9875, "reward/mean_ndcg_at_10": 0.8953125, "reward/mean_r_ndcg": 0.76015625, "reward/mean_r_ndcg_cutoff": 6.86875, "reward/mean_recall": 0.24140625, "reward/mean_recall_at_10": 0.04853515625, "reward/mean_recall_at_100": 0.24140625, "reward/mean_retrieval_combined_bounded": 0.02874755859375, "reward/mean_retrieval_combined_raw": 0.03626241683959961, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1296.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.29476168751716614, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.294921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6241175174713134, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1892630159854889, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.54598789215088, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22549186944961547, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1300, "step_time": 13.7929901610245 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.24609375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.3759170025587082, "epoch": 5.142292490118577, "frac_reward_zero_std": 0.0, "num_tokens": 77676407.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6770008827551245, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7263485001991036, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8963093145869947, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9964850615114236, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6099302195681584, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6648845263473906, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8533333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 1.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.33532639750035315, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3755926004814127, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5445783132530121, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8409638554216867, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6242439836707837, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6335246640636483, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7634343942442534, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9203973843058351, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5719557912027864, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.36341196944132476, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17933667069017675, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.31213525743421705, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 3.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3449.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.41866912653621396, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.45515466850142877, "retrieval_eval/nq/single_turn/recall_at_10": 0.6556248187880545, "retrieval_eval/nq/single_turn/recall_at_100": 0.8958876969169807, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6715372654754367, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7033102398196125, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8408333333333333, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9670000000000001, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0285186767578125, "reward/completion_length_max": 33.0, "reward/completion_length_mean": 30.25, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.91015625, "reward/mean_r_ndcg": 0.84375, "reward/mean_r_ndcg_cutoff": 5.71875, "reward/mean_recall": 0.177734375, "reward/mean_recall_at_10": 0.046875, "reward/mean_recall_at_100": 0.177734375, "reward/mean_retrieval_combined_bounded": 0.032470703125, "reward/mean_retrieval_combined_raw": 0.0283203125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1296.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2669983208179474, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.267578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.706284761428833, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15618681907653809, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.33432674407959, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23345372080802917, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1301, "step_time": 13.61967271193862 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.65625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.25, "completions/min_terminated_length": 0.0, "entropy": 0.3324846597388387, "epoch": 5.158102766798419, "frac_reward_zero_std": 0.0, "grad_norm": 2.4251763820648193, "learning_rate": 2e-05, "loss": -0.00272216796875, "num_tokens": 77914967.0, "reward": 0.999176025390625, "reward/completion_length_max": 38.25, "reward/completion_length_mean": 31.65625, "reward/completion_length_min": 25.25, "reward/completion_length_p90": 34.75, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.001953125, "reward/low_length_query_rate": 0.001953125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.9990234375, "reward/mean_colbert": 0.970703125, "reward/mean_mrr": 0.9951171875, "reward/mean_ndcg_at_10": 0.8798828125, "reward/mean_r_ndcg": 0.748046875, "reward/mean_r_ndcg_cutoff": 6.171875, "reward/mean_recall": 0.19091796875, "reward/mean_recall_at_10": 0.04913330078125, "reward/mean_recall_at_100": 0.19091796875, "reward/mean_retrieval_combined_bounded": 0.0079345703125, "reward/mean_retrieval_combined_raw": -0.07183803804218769, "reward/output_tag_rate": 0.998046875, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1304.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0009765625, "reward/too_short_query_rate": 0.0009765625, "reward_std": 0.31308286637067795, "rewards/retrieval_reward/mean": 0.9990234375, "rewards/retrieval_reward/std": 0.3134765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8952434062957764, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.22046494856476784, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.251892566680908, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2174552083015442, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1305, "step_time": 13.953543272291427 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.6125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.33506149277091024, "epoch": 5.177865612648222, "frac_reward_zero_std": 0.0, "grad_norm": 2.5877208709716797, "learning_rate": 2e-05, "loss": -0.000848388671875, "num_tokens": 78213127.0, "reward": 1.02576904296875, "reward/completion_length_max": 36.2, "reward/completion_length_mean": 31.625, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 33.8, "reward/completion_length_p95": 34.4, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.028125, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.92109375, "reward/mean_r_ndcg": 0.7453125, "reward/mean_r_ndcg_cutoff": 6.65625, "reward/mean_recall": 0.2251953125, "reward/mean_recall_at_10": 0.049609375, "reward/mean_recall_at_100": 0.2251953125, "reward/mean_retrieval_combined_bounded": 0.02882232666015625, "reward/mean_retrieval_combined_raw": 0.09462897181510925, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1307.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28445960879325866, "rewards/retrieval_reward/mean": 1.028125, "rewards/retrieval_reward/std": 0.284765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5164504051208496, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2085232228040695, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.021751594543456, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23412371873855592, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1310, "step_time": 13.800497985177207 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.540625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.3136167511343956, "epoch": 5.1976284584980235, "frac_reward_zero_std": 0.0, "grad_norm": 2.187753200531006, "learning_rate": 2e-05, "loss": -9.8419189453125e-05, "num_tokens": 78511995.0, "reward": 1.0070343017578125, "reward/completion_length_max": 37.6, "reward/completion_length_mean": 32.5, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 34.8, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.0140625, "reward/duplicate_ranked_plain_query_rate": 0.0140625, "reward/invalid_output_rate": 0.0140625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00625, "reward/mean_colbert": 0.96015625, "reward/mean_mrr": 0.9859375, "reward/mean_ndcg_at_10": 0.89921875, "reward/mean_r_ndcg": 0.70703125, "reward/mean_r_ndcg_cutoff": 6.00625, "reward/mean_recall": 0.2126953125, "reward/mean_recall_at_10": 0.047509765625, "reward/mean_recall_at_100": 0.2126953125, "reward/mean_retrieval_combined_bounded": 0.02103271484375, "reward/mean_retrieval_combined_raw": 0.03725571222603321, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9859375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9859375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1312.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.286162793636322, "rewards/retrieval_reward/mean": 1.00625, "rewards/retrieval_reward/std": 0.28671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8611401081085206, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21525731980800628, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 13.715650367736817, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21637285947799684, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1315, "step_time": 13.84904862859985 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.1984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.6, "completions/min_terminated_length": 0.0, "entropy": 0.32606264650821687, "epoch": 5.217391304347826, "frac_reward_zero_std": 0.0, "grad_norm": 1.66469144821167, "learning_rate": 2e-05, "loss": 3.662109375e-05, "num_tokens": 78811961.0, "reward": 1.0301239013671875, "reward/completion_length_max": 41.0, "reward/completion_length_mean": 33.2, "reward/completion_length_min": 29.6, "reward/completion_length_p90": 36.4, "reward/completion_length_p95": 37.2, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.0140625, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.9609375, "reward/mean_mrr": 0.9859375, "reward/mean_ndcg_at_10": 0.89375, "reward/mean_r_ndcg": 0.78984375, "reward/mean_r_ndcg_cutoff": 6.61875, "reward/mean_recall": 0.221875, "reward/mean_recall_at_10": 0.047265625, "reward/mean_recall_at_100": 0.221875, "reward/mean_retrieval_combined_bounded": 0.04599609375, "reward/mean_retrieval_combined_raw": 0.06689453125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9859375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9859375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1316.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00390625, "reward/too_short_query_rate": 0.0, "reward_std": 0.29501203894615174, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.294921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.586194705963135, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2511708527803421, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.521221542358399, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20432668626308442, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1320, "step_time": 14.047670279396698 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 43.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.73828125, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.0, "completions/min_terminated_length": 0.0, "entropy": 0.3360989160835743, "epoch": 5.237154150197629, "frac_reward_zero_std": 0.0, "grad_norm": 2.5312042236328125, "learning_rate": 2e-05, "loss": 0.006494140625, "num_tokens": 79112458.0, "reward": 1.032733154296875, "reward/completion_length_max": 43.0, "reward/completion_length_mean": 33.75, "reward/completion_length_min": 30.0, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.0109375, "reward/length_out_of_range_rate": 0.00625, "reward/low_length_query_rate": 0.00625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0359375, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.9890625, "reward/mean_ndcg_at_10": 0.86796875, "reward/mean_r_ndcg": 0.7796875, "reward/mean_r_ndcg_cutoff": 6.0875, "reward/mean_recall": 0.2646484375, "reward/mean_recall_at_10": 0.047998046875, "reward/mean_recall_at_100": 0.2646484375, "reward/mean_retrieval_combined_bounded": 0.04677734375, "reward/mean_retrieval_combined_raw": 0.08922290802001953, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1321.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00625, "reward/too_short_query_rate": 0.0, "reward_std": 0.29028287529945374, "rewards/retrieval_reward/mean": 1.0359375, "rewards/retrieval_reward/std": 0.290234375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.3523833751678467, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1863381713628769, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.94567050933838, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22455111443996428, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1325, "step_time": 14.195352535368874 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.92734375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.8, "completions/min_terminated_length": 0.0, "entropy": 0.3399405062198639, "epoch": 5.256916996047431, "frac_reward_zero_std": 0.0, "grad_norm": 2.1887800693511963, "learning_rate": 2e-05, "loss": 0.00033111572265625, "num_tokens": 79412237.0, "reward": 1.032586669921875, "reward/completion_length_max": 40.8, "reward/completion_length_mean": 32.95, "reward/completion_length_min": 27.8, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.01484375, "reward/duplicate_ranked_plain_query_rate": 0.01484375, "reward/invalid_output_rate": 0.01796875, "reward/length_out_of_range_rate": 0.003125, "reward/low_length_query_rate": 0.003125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.034375, "reward/mean_colbert": 0.959375, "reward/mean_mrr": 0.98203125, "reward/mean_ndcg_at_10": 0.90546875, "reward/mean_r_ndcg": 0.80390625, "reward/mean_r_ndcg_cutoff": 5.78125, "reward/mean_recall": 0.2162109375, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.2162109375, "reward/mean_retrieval_combined_bounded": 0.0521240234375, "reward/mean_retrieval_combined_raw": 0.081298828125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98203125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98203125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1328.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.2979713618755341, "rewards/retrieval_reward/mean": 1.034375, "rewards/retrieval_reward/std": 0.298828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.916709041595459, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.23244471251964569, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 13.352948188781738, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2315653681755066, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1330, "step_time": 14.016945830010808 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.0171875, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.30841902568936347, "epoch": 5.276679841897233, "frac_reward_zero_std": 0.0, "grad_norm": 2.2072973251342773, "learning_rate": 2e-05, "loss": 0.00260009765625, "num_tokens": 79711683.0, "reward": 1.0365203857421874, "reward/completion_length_max": 41.2, "reward/completion_length_mean": 32.95, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.0109375, "reward/duplicate_ranked_plain_query_rate": 0.0109375, "reward/invalid_output_rate": 0.015625, "reward/length_out_of_range_rate": 0.0046875, "reward/low_length_query_rate": 0.0046875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0375, "reward/mean_colbert": 0.96171875, "reward/mean_mrr": 0.984375, "reward/mean_ndcg_at_10": 0.90859375, "reward/mean_r_ndcg": 0.79921875, "reward/mean_r_ndcg_cutoff": 7.1375, "reward/mean_recall": 0.2572265625, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.2572265625, "reward/mean_retrieval_combined_bounded": 0.05439453125, "reward/mean_retrieval_combined_raw": 0.0685791015625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1331.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0046875, "reward/too_short_query_rate": 0.0, "reward_std": 0.3098415911197662, "rewards/retrieval_reward/mean": 1.0375, "rewards/retrieval_reward/std": 0.310546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8045578956604005, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2502133220434189, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.821302604675292, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21784914135932923, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1335, "step_time": 14.023873454413842 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.1203125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.2, "completions/min_terminated_length": 0.0, "entropy": 0.32110289856791496, "epoch": 5.296442687747035, "frac_reward_zero_std": 0.0, "grad_norm": 2.062063694000244, "learning_rate": 2e-05, "loss": 0.00203857421875, "num_tokens": 80011709.0, "reward": 1.0194549560546875, "reward/completion_length_max": 38.8, "reward/completion_length_mean": 33.1, "reward/completion_length_min": 29.2, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 36.2, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.0140625, "reward/length_out_of_range_rate": 0.00703125, "reward/low_length_query_rate": 0.00703125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.96171875, "reward/mean_mrr": 0.9859375, "reward/mean_ndcg_at_10": 0.8890625, "reward/mean_r_ndcg": 0.78671875, "reward/mean_r_ndcg_cutoff": 5.4625, "reward/mean_recall": 0.2466796875, "reward/mean_recall_at_10": 0.048291015625, "reward/mean_recall_at_100": 0.2466796875, "reward/mean_retrieval_combined_bounded": 0.037030029296875, "reward/mean_retrieval_combined_raw": 0.05248995237052441, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9859375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9859375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1336.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00703125, "reward/too_short_query_rate": 0.0, "reward_std": 0.30835007429122924, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.30859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.712549829483032, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18121292740106582, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.638926887512207, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22789108157157897, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1340, "step_time": 13.962606554047671 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.07109375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.36328420266509054, "epoch": 5.316205533596838, "frac_reward_zero_std": 0.0, "grad_norm": 2.1742656230926514, "learning_rate": 2e-05, "loss": 3.871917724609375e-05, "num_tokens": 80311384.0, "reward": 1.013739013671875, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 33.05, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 36.2, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.9265625, "reward/mean_r_ndcg": 0.81640625, "reward/mean_r_ndcg_cutoff": 6.38125, "reward/mean_recall": 0.2390625, "reward/mean_recall_at_10": 0.049365234375, "reward/mean_recall_at_100": 0.2390625, "reward/mean_retrieval_combined_bounded": 0.0176513671875, "reward/mean_retrieval_combined_raw": 0.018262859620153903, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1344.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28287009000778196, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.283203125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.75417160987854, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16385654509067535, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.98508186340332, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25436362326145173, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1345, "step_time": 13.942598337004892 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.63046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.8, "completions/min_terminated_length": 0.0, "entropy": 0.3478271380066872, "epoch": 5.335968379446641, "frac_reward_zero_std": 0.0, "grad_norm": 2.244884729385376, "learning_rate": 2e-05, "loss": -6.9427490234375e-05, "num_tokens": 80612351.0, "reward": 1.033026123046875, "reward/completion_length_max": 39.4, "reward/completion_length_mean": 33.55, "reward/completion_length_min": 29.8, "reward/completion_length_p90": 36.4, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.9015625, "reward/mean_r_ndcg": 0.70859375, "reward/mean_r_ndcg_cutoff": 5.0375, "reward/mean_recall": 0.206640625, "reward/mean_recall_at_10": 0.04921875, "reward/mean_recall_at_100": 0.206640625, "reward/mean_retrieval_combined_bounded": 0.03797607421875, "reward/mean_retrieval_combined_raw": 0.06092561259865761, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1345.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.27722388207912446, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.27734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.820400285720825, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16580475270748138, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.694703292846679, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25272119641304014, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1350, "step_time": 14.023041232232936 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.0859375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.2975800335407257, "epoch": 5.339920948616601, "frac_reward_zero_std": 0.0, "num_tokens": 80672533.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6788553790062314, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7295917404913791, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9024604569420035, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9973637961335676, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5988900629245334, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.656733472512674, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8533333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.32788389001160684, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.366542175774221, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5325301204819277, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8409638554216867, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6267108864972728, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6401183298378231, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7679798487897079, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9085135814889337, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.564313506578611, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.35637984043949905, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.174075233778379, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.31238050472105006, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 5.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3447.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.4087882411348722, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.44653798182752596, "retrieval_eval/nq/single_turn/recall_at_10": 0.650299777584373, "retrieval_eval/nq/single_turn/recall_at_100": 0.8886954839957452, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6702029410486742, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7012301476777077, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8356666666666667, "retrieval_eval/scifact/single_turn/recall_at_100": 0.96, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.022308349609375, "reward/completion_length_max": 39.0, "reward/completion_length_mean": 34.0, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.98046875, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.96484375, "reward/mean_r_ndcg": 0.9140625, "reward/mean_r_ndcg_cutoff": 8.25, "reward/mean_recall": 0.2490234375, "reward/mean_recall_at_10": 0.0498046875, "reward/mean_recall_at_100": 0.2490234375, "reward/mean_retrieval_combined_bounded": 0.0220947265625, "reward/mean_retrieval_combined_raw": 3.314018249511719e-05, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1352.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2553945779800415, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.255859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7097878456115723, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.24174797534942627, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.976968765258789, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2198600471019745, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1351, "step_time": 14.047966452024411 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.7099609375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.32621996477246284, "epoch": 5.355731225296442, "frac_reward_zero_std": 0.0, "grad_norm": 1.744583010673523, "learning_rate": 2e-05, "loss": 0.0012451171875, "num_tokens": 80912780.0, "reward": 1.0283584594726562, "reward/completion_length_max": 40.25, "reward/completion_length_mean": 33.6875, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 36.25, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.0029296875, "reward/duplicate_ranked_plain_query_rate": 0.0029296875, "reward/invalid_output_rate": 0.0107421875, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.00390625, "reward/malformed_ranked_plain_query_rate": 0.00390625, "reward/mean": 1.029296875, "reward/mean_colbert": 0.966796875, "reward/mean_mrr": 0.9892578125, "reward/mean_ndcg_at_10": 0.8974609375, "reward/mean_r_ndcg": 0.8447265625, "reward/mean_r_ndcg_cutoff": 7.640625, "reward/mean_recall": 0.243896484375, "reward/mean_recall_at_10": 0.04803466796875, "reward/mean_recall_at_100": 0.243896484375, "reward/mean_retrieval_combined_bounded": 0.04486083984375, "reward/mean_retrieval_combined_raw": 0.06573486328125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9892578125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9892578125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1352.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00390625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3066411465406418, "rewards/retrieval_reward/mean": 1.029296875, "rewards/retrieval_reward/std": 0.30712890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8164848685264587, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1718487460166216, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.220872402191162, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24023648723959923, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1355, "step_time": 13.974543632764835 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 59.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.675, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.4, "completions/min_terminated_length": 0.0, "entropy": 0.34757907316088676, "epoch": 5.375494071146245, "frac_reward_zero_std": 0.0, "grad_norm": 1.9367358684539795, "learning_rate": 2e-05, "loss": 0.01611328125, "num_tokens": 81213612.0, "reward": 1.03631591796875, "reward/completion_length_max": 59.8, "reward/completion_length_mean": 33.7, "reward/completion_length_min": 29.4, "reward/completion_length_p90": 37.0, "reward/completion_length_p95": 38.4, "reward/duplicate_query_rate": 0.0140625, "reward/duplicate_ranked_plain_query_rate": 0.0140625, "reward/invalid_output_rate": 0.028125, "reward/length_out_of_range_rate": 0.01328125, "reward/low_length_query_rate": 0.01328125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0375, "reward/mean_colbert": 0.9484375, "reward/mean_mrr": 0.971875, "reward/mean_ndcg_at_10": 0.8765625, "reward/mean_r_ndcg": 0.77421875, "reward/mean_r_ndcg_cutoff": 6.975, "reward/mean_recall": 0.2236328125, "reward/mean_recall_at_10": 0.04736328125, "reward/mean_recall_at_100": 0.2236328125, "reward/mean_retrieval_combined_bounded": 0.072509765625, "reward/mean_retrieval_combined_raw": 0.1896484375, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.971875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.971875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1358.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01328125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3419324517250061, "rewards/retrieval_reward/mean": 1.0375, "rewards/retrieval_reward/std": 0.34296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6642000675201416, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1739524781703949, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.894085121154784, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2514154314994812, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1360, "step_time": 14.903630025021267 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.609375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.8, "completions/min_terminated_length": 0.0, "entropy": 0.35526270866394044, "epoch": 5.395256916996048, "frac_reward_zero_std": 0.0, "grad_norm": 2.1335248947143555, "learning_rate": 2e-05, "loss": -0.001177978515625, "num_tokens": 81512920.0, "reward": 1.0095977783203125, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 32.625, "reward/completion_length_min": 28.8, "reward/completion_length_p90": 35.4, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.0203125, "reward/duplicate_ranked_plain_query_rate": 0.0203125, "reward/invalid_output_rate": 0.02109375, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00859375, "reward/mean_colbert": 0.9546875, "reward/mean_mrr": 0.97890625, "reward/mean_ndcg_at_10": 0.890625, "reward/mean_r_ndcg": 0.7625, "reward/mean_r_ndcg_cutoff": 6.01875, "reward/mean_recall": 0.21953125, "reward/mean_recall_at_10": 0.047509765625, "reward/mean_recall_at_100": 0.21953125, "reward/mean_retrieval_combined_bounded": 0.03126220703125, "reward/mean_retrieval_combined_raw": 0.04977983236312866, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1360.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2986106753349304, "rewards/retrieval_reward/mean": 1.00859375, "rewards/retrieval_reward/std": 0.298046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.778883695602417, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.19303802251815796, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.956507015228272, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24151320159435272, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1365, "step_time": 13.910969562991522 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.3421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.37710077315568924, "epoch": 5.41501976284585, "frac_reward_zero_std": 0.0, "grad_norm": 2.25579571723938, "learning_rate": 2e-05, "loss": 0.001373291015625, "num_tokens": 81811950.0, "reward": 1.01002197265625, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 32.325, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.01484375, "reward/duplicate_ranked_plain_query_rate": 0.01484375, "reward/invalid_output_rate": 0.015625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00859375, "reward/mean_colbert": 0.95859375, "reward/mean_mrr": 0.984375, "reward/mean_ndcg_at_10": 0.865625, "reward/mean_r_ndcg": 0.68671875, "reward/mean_r_ndcg_cutoff": 5.25, "reward/mean_recall": 0.2263671875, "reward/mean_recall_at_10": 0.047216796875, "reward/mean_recall_at_100": 0.2263671875, "reward/mean_retrieval_combined_bounded": 0.026104736328125, "reward/mean_retrieval_combined_raw": 0.05395543277263641, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1368.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.29498420357704164, "rewards/retrieval_reward/mean": 1.00859375, "rewards/retrieval_reward/std": 0.2953125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8808106422424316, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.175565505027771, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 15.423621368408202, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25937082171440123, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1370, "step_time": 13.896243293222506 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 55.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.1203125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.3654010213911533, "epoch": 5.434782608695652, "frac_reward_zero_std": 0.0, "grad_norm": 2.0375916957855225, "learning_rate": 2e-05, "loss": 0.01201171875, "num_tokens": 82110664.0, "reward": 1.03280029296875, "reward/completion_length_max": 55.0, "reward/completion_length_mean": 32.125, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.00546875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0328125, "reward/mean_colbert": 0.9703125, "reward/mean_mrr": 0.99453125, "reward/mean_ndcg_at_10": 0.91328125, "reward/mean_r_ndcg": 0.790625, "reward/mean_r_ndcg_cutoff": 6.2875, "reward/mean_recall": 0.22890625, "reward/mean_recall_at_10": 0.048193359375, "reward/mean_recall_at_100": 0.22890625, "reward/mean_retrieval_combined_bounded": 0.0400146484375, "reward/mean_retrieval_combined_raw": 0.1116943359375, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.99453125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99453125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1372.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2975848317146301, "rewards/retrieval_reward/mean": 1.0328125, "rewards/retrieval_reward/std": 0.2978515625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.707711696624756, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2083366960287094, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.45725040435791, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24301610589027406, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1375, "step_time": 14.649206547194627 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 57.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.3375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.36145311072468755, "epoch": 5.454545454545454, "frac_reward_zero_std": 0.0, "grad_norm": 2.022620677947998, "learning_rate": 2e-05, "loss": 0.012158203125, "num_tokens": 82409592.0, "reward": 1.017767333984375, "reward/completion_length_max": 57.4, "reward/completion_length_mean": 32.3, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 35.4, "reward/completion_length_p95": 36.2, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.01875, "reward/length_out_of_range_rate": 0.01015625, "reward/low_length_query_rate": 0.01015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01796875, "reward/mean_colbert": 0.9578125, "reward/mean_mrr": 0.98125, "reward/mean_ndcg_at_10": 0.88359375, "reward/mean_r_ndcg": 0.790625, "reward/mean_r_ndcg_cutoff": 6.5625, "reward/mean_recall": 0.2306640625, "reward/mean_recall_at_10": 0.048486328125, "reward/mean_recall_at_100": 0.2306640625, "reward/mean_retrieval_combined_bounded": 0.0431884765625, "reward/mean_retrieval_combined_raw": 0.0740966796875, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.98125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1376.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3182853996753693, "rewards/retrieval_reward/mean": 1.01796875, "rewards/retrieval_reward/std": 0.319140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.832656669616699, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18708837628364564, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.445702171325683, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23753597736358642, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1380, "step_time": 14.792433609836735 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.47890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.3461582988500595, "epoch": 5.474308300395257, "frac_reward_zero_std": 0.0, "grad_norm": 1.9715555906295776, "learning_rate": 2e-05, "loss": -0.001177978515625, "num_tokens": 82707645.0, "reward": 1.0266693115234375, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 31.5, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 34.6, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.9703125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.89609375, "reward/mean_r_ndcg": 0.81875, "reward/mean_r_ndcg_cutoff": 6.325, "reward/mean_recall": 0.2181640625, "reward/mean_recall_at_10": 0.04853515625, "reward/mean_recall_at_100": 0.2181640625, "reward/mean_retrieval_combined_bounded": 0.03336181640625, "reward/mean_retrieval_combined_raw": 0.045751953125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1382.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.27528536319732666, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.27578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.66348614692688, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.23456915020942687, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.158835411071777, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21190106570720674, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1385, "step_time": 13.861884392204228 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.87890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.6, "completions/min_terminated_length": 0.0, "entropy": 0.3362236678600311, "epoch": 5.4940711462450595, "frac_reward_zero_std": 0.0, "grad_norm": 1.9871816635131836, "learning_rate": 2e-05, "loss": -0.000921630859375, "num_tokens": 83004482.0, "reward": 1.03253173828125, "reward/completion_length_max": 35.4, "reward/completion_length_mean": 30.9, "reward/completion_length_min": 26.6, "reward/completion_length_p90": 32.8, "reward/completion_length_p95": 33.2, "reward/duplicate_query_rate": 0.01875, "reward/duplicate_ranked_plain_query_rate": 0.01875, "reward/invalid_output_rate": 0.01875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0328125, "reward/mean_colbert": 0.95625, "reward/mean_mrr": 0.98125, "reward/mean_ndcg_at_10": 0.90859375, "reward/mean_r_ndcg": 0.76015625, "reward/mean_r_ndcg_cutoff": 7.1125, "reward/mean_recall": 0.221484375, "reward/mean_recall_at_10": 0.048388671875, "reward/mean_recall_at_100": 0.221484375, "reward/mean_retrieval_combined_bounded": 0.05150146484375, "reward/mean_retrieval_combined_raw": 0.08710889220237732, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1387.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2972429871559143, "rewards/retrieval_reward/mean": 1.0328125, "rewards/retrieval_reward/std": 0.297265625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6843995094299316, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20426426231861114, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.459609794616699, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22131688594818116, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1390, "step_time": 13.714245091413613 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 74.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.60859375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.8, "completions/min_terminated_length": 0.0, "entropy": 0.33917436860501765, "epoch": 5.513833992094861, "frac_reward_zero_std": 0.0, "grad_norm": 2.3794658184051514, "learning_rate": 2e-05, "loss": 0.044921875, "num_tokens": 83303597.0, "reward": 0.970172119140625, "reward/completion_length_max": 74.6, "reward/completion_length_mean": 32.65, "reward/completion_length_min": 27.8, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 52.8, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.03359375, "reward/length_out_of_range_rate": 0.0171875, "reward/low_length_query_rate": 0.0171875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.9703125, "reward/mean_colbert": 0.94296875, "reward/mean_mrr": 0.96640625, "reward/mean_ndcg_at_10": 0.87109375, "reward/mean_r_ndcg": 0.725, "reward/mean_r_ndcg_cutoff": 5.71875, "reward/mean_recall": 0.223828125, "reward/mean_recall_at_10": 0.04677734375, "reward/mean_recall_at_100": 0.223828125, "reward/mean_retrieval_combined_bounded": 0.03420677185058594, "reward/mean_retrieval_combined_raw": 0.0716064453125, "reward/output_tag_rate": 0.9890625, "reward/query_valid_rate": 0.96640625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96640625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1392.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0171875, "reward/too_short_query_rate": 0.0, "reward_std": 0.41144264340400694, "rewards/retrieval_reward/mean": 0.9703125, "rewards/retrieval_reward/std": 0.411328125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.869975233078003, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2471234381198883, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.81668529510498, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20554466247558595, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1395, "step_time": 15.578041481401305 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.72109375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.32787725850939753, "epoch": 5.533596837944664, "frac_reward_zero_std": 0.0, "grad_norm": 2.6320879459381104, "learning_rate": 2e-05, "loss": 0.001080322265625, "num_tokens": 83601864.0, "reward": 1.0195587158203125, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 31.725, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 34.8, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.8890625, "reward/mean_r_ndcg": 0.734375, "reward/mean_r_ndcg_cutoff": 5.60625, "reward/mean_recall": 0.240234375, "reward/mean_recall_at_10": 0.0484375, "reward/mean_recall_at_100": 0.240234375, "reward/mean_retrieval_combined_bounded": 0.02391357421875, "reward/mean_retrieval_combined_raw": 0.026494288444519044, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1396.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.29383531808853147, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.29453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6865302085876466, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20923213064670562, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 13.343101692199706, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22663134336471558, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1400, "step_time": 13.896238592185545 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.30640074983239174, "epoch": 5.537549407114625, "frac_reward_zero_std": 0.0, "num_tokens": 83661612.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6674900327380312, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7186816662659106, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8954305799648506, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9973637961335676, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6203585512785429, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6688315695820463, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.84, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.34026321275640886, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.380040629942726, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5469879518072289, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8457831325301205, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6115312959927877, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6288182510083308, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7722749375038107, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9133913328455582, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5791478343657935, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.370112551978492, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.18470619536710592, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3212878459725756, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 4.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3448.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.42441722234815454, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.45956898714641686, "retrieval_eval/nq/single_turn/recall_at_10": 0.6571200696055685, "retrieval_eval/nq/single_turn/recall_at_100": 0.8945523975251353, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6779400001227484, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7131155869880776, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8631666666666666, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9460000000000001, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.03192138671875, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 31.75, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.01953125, "reward/duplicate_ranked_plain_query_rate": 0.01953125, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.953125, "reward/mean_mrr": 0.98046875, "reward/mean_ndcg_at_10": 0.84765625, "reward/mean_r_ndcg": 0.68359375, "reward/mean_r_ndcg_cutoff": 3.90625, "reward/mean_recall": 0.255859375, "reward/mean_recall_at_10": 0.047119140625, "reward/mean_recall_at_100": 0.255859375, "reward/mean_retrieval_combined_bounded": 0.051025390625, "reward/mean_retrieval_combined_raw": 0.0810546875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1400.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.268261194229126, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.26953125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6025216579437256, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1792631447315216, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.32207202911377, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20250631868839264, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1401, "step_time": 13.833392393833492 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.9169921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.25, "completions/min_terminated_length": 0.0, "entropy": 0.37754093389958143, "epoch": 5.553359683794467, "frac_reward_zero_std": 0.0, "grad_norm": 2.0141854286193848, "learning_rate": 2e-05, "loss": -0.0007354736328125, "num_tokens": 83899959.0, "reward": 0.9894256591796875, "reward/completion_length_max": 36.25, "reward/completion_length_mean": 31.90625, "reward/completion_length_min": 28.25, "reward/completion_length_p90": 34.25, "reward/completion_length_p95": 34.25, "reward/duplicate_query_rate": 0.03515625, "reward/duplicate_ranked_plain_query_rate": 0.03515625, "reward/invalid_output_rate": 0.03515625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.9873046875, "reward/mean_colbert": 0.9423828125, "reward/mean_mrr": 0.9638671875, "reward/mean_ndcg_at_10": 0.8662109375, "reward/mean_r_ndcg": 0.77734375, "reward/mean_r_ndcg_cutoff": 5.984375, "reward/mean_recall": 0.23291015625, "reward/mean_recall_at_10": 0.04779052734375, "reward/mean_recall_at_100": 0.23291015625, "reward/mean_retrieval_combined_bounded": 0.0246124267578125, "reward/mean_retrieval_combined_raw": 0.03729332983493805, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.96484375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96484375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1402.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3314291760325432, "rewards/retrieval_reward/mean": 0.9873046875, "rewards/retrieval_reward/std": 0.33154296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.645044207572937, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2046031802892685, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.218164443969727, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23260925337672234, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1405, "step_time": 13.76417278846202 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.428125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.8, "completions/min_terminated_length": 0.0, "entropy": 0.3623616650700569, "epoch": 5.573122529644269, "frac_reward_zero_std": 0.0, "grad_norm": 2.601083517074585, "learning_rate": 2e-05, "loss": 0.00269775390625, "num_tokens": 84198843.0, "reward": 1.0260162353515625, "reward/completion_length_max": 39.6, "reward/completion_length_mean": 32.425, "reward/completion_length_min": 27.8, "reward/completion_length_p90": 35.4, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.0109375, "reward/duplicate_ranked_plain_query_rate": 0.0109375, "reward/invalid_output_rate": 0.01484375, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.96015625, "reward/mean_mrr": 0.98515625, "reward/mean_ndcg_at_10": 0.90625, "reward/mean_r_ndcg": 0.79609375, "reward/mean_r_ndcg_cutoff": 5.88125, "reward/mean_recall": 0.2232421875, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.2232421875, "reward/mean_retrieval_combined_bounded": 0.0427490234375, "reward/mean_retrieval_combined_raw": 0.07760963439941407, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98515625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98515625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1408.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00390625, "reward/too_short_query_rate": 0.0, "reward_std": 0.2986110270023346, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.299609375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.783676815032959, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.23464515805244446, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.105033493041992, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22093598246574403, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1410, "step_time": 13.921360773406922 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.33046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.2, "completions/min_terminated_length": 0.0, "entropy": 0.33970851004123687, "epoch": 5.592885375494071, "frac_reward_zero_std": 0.0, "grad_norm": 2.687227249145508, "learning_rate": 2e-05, "loss": -0.00108642578125, "num_tokens": 84498018.0, "reward": 1.008099365234375, "reward/completion_length_max": 38.2, "reward/completion_length_mean": 32.325, "reward/completion_length_min": 29.2, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00859375, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.87265625, "reward/mean_r_ndcg": 0.7375, "reward/mean_r_ndcg_cutoff": 6.525, "reward/mean_recall": 0.2212890625, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.2212890625, "reward/mean_retrieval_combined_bounded": 0.0096435546875, "reward/mean_retrieval_combined_raw": -0.06679904460906982, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1411.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2871101498603821, "rewards/retrieval_reward/mean": 1.00859375, "rewards/retrieval_reward/std": 0.2875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8011523246765138, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.22797303199768065, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.687436294555663, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23187375664711, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1415, "step_time": 13.906900051573757 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.2640625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.35609782114624977, "epoch": 5.612648221343873, "frac_reward_zero_std": 0.0, "grad_norm": 1.9243884086608887, "learning_rate": 2e-05, "loss": 0.000811767578125, "num_tokens": 84796884.0, "reward": 1.0109527587890625, "reward/completion_length_max": 37.6, "reward/completion_length_mean": 32.25, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 34.8, "reward/completion_length_p95": 36.2, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.009375, "reward/low_length_query_rate": 0.009375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0109375, "reward/mean_colbert": 0.95546875, "reward/mean_mrr": 0.98046875, "reward/mean_ndcg_at_10": 0.871875, "reward/mean_r_ndcg": 0.76796875, "reward/mean_r_ndcg_cutoff": 6.10625, "reward/mean_recall": 0.229296875, "reward/mean_recall_at_10": 0.04755859375, "reward/mean_recall_at_100": 0.229296875, "reward/mean_retrieval_combined_bounded": 0.03515625, "reward/mean_retrieval_combined_raw": 0.052685213088989255, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1416.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.009375, "reward/too_short_query_rate": 0.0, "reward_std": 0.31483805775642393, "rewards/retrieval_reward/mean": 1.0109375, "rewards/retrieval_reward/std": 0.315625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8664306640625, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2138090491294861, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.313171577453613, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2357971340417862, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1420, "step_time": 13.863184759148862 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.090625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.3371045172214508, "epoch": 5.632411067193676, "frac_reward_zero_std": 0.0, "grad_norm": 2.0914247035980225, "learning_rate": 2e-05, "loss": 0.0005401611328125, "num_tokens": 85097032.0, "reward": 1.0194091796875, "reward/completion_length_max": 38.4, "reward/completion_length_mean": 33.05, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.00234375, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.99453125, "reward/mean_ndcg_at_10": 0.8734375, "reward/mean_r_ndcg": 0.76484375, "reward/mean_r_ndcg_cutoff": 5.9125, "reward/mean_recall": 0.2419921875, "reward/mean_recall_at_10": 0.04853515625, "reward/mean_recall_at_100": 0.2419921875, "reward/mean_retrieval_combined_bounded": 0.022003173828125, "reward/mean_retrieval_combined_raw": 0.017431437969207764, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1424.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.27053308486938477, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.270703125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.685869741439819, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20776941180229186, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.766100120544433, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2274033546447754, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1425, "step_time": 13.881838974251878 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.6625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.4, "completions/min_terminated_length": 0.0, "entropy": 0.39293897449970244, "epoch": 5.6521739130434785, "frac_reward_zero_std": 0.0, "grad_norm": 1.7946759462356567, "learning_rate": 2e-05, "loss": 0.0005645751953125, "num_tokens": 85397624.0, "reward": 1.0191314697265625, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 33.65, "reward/completion_length_min": 30.4, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.2, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01953125, "reward/mean_colbert": 0.9671875, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.896875, "reward/mean_r_ndcg": 0.80078125, "reward/mean_r_ndcg_cutoff": 6.7125, "reward/mean_recall": 0.2359375, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.2359375, "reward/mean_retrieval_combined_bounded": 0.0262359619140625, "reward/mean_retrieval_combined_raw": 0.03315112590789795, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1425.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3017019033432007, "rewards/retrieval_reward/mean": 1.01953125, "rewards/retrieval_reward/std": 0.301953125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.459931230545044, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.152432356774807, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.272514152526856, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25455255806446075, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1430, "step_time": 13.903074617637321 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 35.090625, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.6, "completions/min_terminated_length": 0.0, "entropy": 0.3776115469634533, "epoch": 5.67193675889328, "frac_reward_zero_std": 0.0, "grad_norm": 2.378952741622925, "learning_rate": 2e-05, "loss": 0.000537109375, "num_tokens": 85700684.0, "reward": 1.027471923828125, "reward/completion_length_max": 40.4, "reward/completion_length_mean": 35.15, "reward/completion_length_min": 31.6, "reward/completion_length_p90": 37.0, "reward/completion_length_p95": 37.8, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00859375, "reward/length_out_of_range_rate": 0.0046875, "reward/low_length_query_rate": 0.0046875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.965625, "reward/mean_mrr": 0.99140625, "reward/mean_ndcg_at_10": 0.890625, "reward/mean_r_ndcg": 0.76171875, "reward/mean_r_ndcg_cutoff": 6.16875, "reward/mean_recall": 0.2326171875, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.2326171875, "reward/mean_retrieval_combined_bounded": 0.03839111328125, "reward/mean_retrieval_combined_raw": 0.057568359375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99140625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99140625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1432.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0046875, "reward/too_short_query_rate": 0.0, "reward_std": 0.2855036616325378, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.286328125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6571630954742433, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16145142912864685, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.95568027496338, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24975697994232177, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1435, "step_time": 14.097280194621998 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.12421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.4, "completions/min_terminated_length": 0.0, "entropy": 0.4014681227505207, "epoch": 5.691699604743083, "frac_reward_zero_std": 0.0, "grad_norm": 2.3131041526794434, "learning_rate": 2e-05, "loss": 0.00196533203125, "num_tokens": 86004331.0, "reward": 1.004736328125, "reward/completion_length_max": 41.4, "reward/completion_length_mean": 36.1, "reward/completion_length_min": 32.4, "reward/completion_length_p90": 38.0, "reward/completion_length_p95": 38.8, "reward/duplicate_query_rate": 0.01953125, "reward/duplicate_ranked_plain_query_rate": 0.01953125, "reward/invalid_output_rate": 0.0328125, "reward/length_out_of_range_rate": 0.01328125, "reward/low_length_query_rate": 0.01328125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0046875, "reward/mean_colbert": 0.94453125, "reward/mean_mrr": 0.9671875, "reward/mean_ndcg_at_10": 0.890625, "reward/mean_r_ndcg": 0.7734375, "reward/mean_r_ndcg_cutoff": 5.75625, "reward/mean_recall": 0.2140625, "reward/mean_recall_at_10": 0.0474609375, "reward/mean_recall_at_100": 0.2140625, "reward/mean_retrieval_combined_bounded": 0.0441162109375, "reward/mean_retrieval_combined_raw": 0.09004029035568237, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9671875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9671875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1438.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01328125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3441204190254211, "rewards/retrieval_reward/mean": 1.0046875, "rewards/retrieval_reward/std": 0.344921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.501054120063782, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13258386626839638, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.722441482543946, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2765680402517319, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1440, "step_time": 14.05633911433397 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.8, "completions/mean_terminated_length": 0.0, "completions/min_length": 33.0, "completions/min_terminated_length": 0.0, "entropy": 0.42963493466377256, "epoch": 5.711462450592886, "frac_reward_zero_std": 0.0, "grad_norm": 1.7337027788162231, "learning_rate": 2e-05, "loss": 0.002587890625, "num_tokens": 86308875.0, "reward": 1.0076080322265626, "reward/completion_length_max": 41.2, "reward/completion_length_mean": 36.8, "reward/completion_length_min": 33.0, "reward/completion_length_p90": 38.8, "reward/completion_length_p95": 39.0, "reward/duplicate_query_rate": 0.01640625, "reward/duplicate_ranked_plain_query_rate": 0.01640625, "reward/invalid_output_rate": 0.0421875, "reward/length_out_of_range_rate": 0.02578125, "reward/low_length_query_rate": 0.02578125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00703125, "reward/mean_colbert": 0.9328125, "reward/mean_mrr": 0.9578125, "reward/mean_ndcg_at_10": 0.88046875, "reward/mean_r_ndcg": 0.8, "reward/mean_r_ndcg_cutoff": 6.575, "reward/mean_recall": 0.23671875, "reward/mean_recall_at_10": 0.04755859375, "reward/mean_recall_at_100": 0.23671875, "reward/mean_retrieval_combined_bounded": 0.062646484375, "reward/mean_retrieval_combined_raw": 0.14501953125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9578125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9578125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1440.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.02578125, "reward/too_short_query_rate": 0.0, "reward_std": 0.34385876059532167, "rewards/retrieval_reward/mean": 1.00703125, "rewards/retrieval_reward/std": 0.34375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6173198223114014, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15259220749139785, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.209788131713868, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2587874621152878, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1445, "step_time": 14.029907160415314 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.14453125, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.8, "completions/min_terminated_length": 0.0, "entropy": 0.4102263882756233, "epoch": 5.7312252964426875, "frac_reward_zero_std": 0.0, "grad_norm": 2.1409802436828613, "learning_rate": 2e-05, "loss": 0.00263671875, "num_tokens": 86612516.0, "reward": 1.0152862548828125, "reward/completion_length_max": 40.8, "reward/completion_length_mean": 36.15, "reward/completion_length_min": 32.8, "reward/completion_length_p90": 38.0, "reward/completion_length_p95": 38.6, "reward/duplicate_query_rate": 0.01171875, "reward/duplicate_ranked_plain_query_rate": 0.01171875, "reward/invalid_output_rate": 0.0203125, "reward/length_out_of_range_rate": 0.00859375, "reward/low_length_query_rate": 0.00859375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01484375, "reward/mean_colbert": 0.95625, "reward/mean_mrr": 0.9796875, "reward/mean_ndcg_at_10": 0.9, "reward/mean_r_ndcg": 0.7703125, "reward/mean_r_ndcg_cutoff": 6.3125, "reward/mean_recall": 0.2490234375, "reward/mean_recall_at_10": 0.048486328125, "reward/mean_recall_at_100": 0.2490234375, "reward/mean_retrieval_combined_bounded": 0.039874267578125, "reward/mean_retrieval_combined_raw": 0.08076147884130477, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9796875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9796875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1448.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00859375, "reward/too_short_query_rate": 0.0, "reward_std": 0.32551525831222533, "rewards/retrieval_reward/mean": 1.01484375, "rewards/retrieval_reward/std": 0.326953125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.635607433319092, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.17612124979496002, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.375314331054687, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2437127947807312, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1450, "step_time": 14.075175356969703 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 42.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 35.484375, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.0, "completions/min_terminated_length": 0.0, "entropy": 0.3895851597189903, "epoch": 5.735177865612648, "frac_reward_zero_std": 0.0, "num_tokens": 86673056.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6521423539954193, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7077516996503622, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.8971880492091389, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9929701230228472, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6145413944197293, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6611169241929881, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8266666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3255081035659455, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3592301067509341, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5132530120481927, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8602409638554217, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6233991025154597, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6373446679765862, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7741669715261265, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9167295439302482, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5846788797256539, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3632787208328561, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17721241925767925, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.32000459820616106, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 4.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3448.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.423975941916059, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4607011185922248, "retrieval_eval/nq/single_turn/recall_at_10": 0.659416086620263, "retrieval_eval/nq/single_turn/recall_at_100": 0.8931747873163186, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6736638445234829, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7063021410601728, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8423333333333333, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9483333333333334, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0147857666015625, "reward/completion_length_max": 42.0, "reward/completion_length_mean": 35.5, "reward/completion_length_min": 32.0, "reward/completion_length_p90": 38.0, "reward/completion_length_p95": 38.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.90625, "reward/mean_r_ndcg": 0.7109375, "reward/mean_r_ndcg_cutoff": 6.9375, "reward/mean_recall": 0.2236328125, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.2236328125, "reward/mean_retrieval_combined_bounded": 0.0203857421875, "reward/mean_retrieval_combined_raw": 0.03173828125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1448.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00390625, "reward/too_short_query_rate": 0.0, "reward_std": 0.2774054706096649, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.27734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.978696584701538, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.26209086179733276, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.529464721679688, "sampling/raw_vllm_sampling_logp_difference/mean": 0.19060736894607544, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1451, "step_time": 14.13536891102558 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 35.69140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.25, "completions/min_terminated_length": 0.0, "entropy": 0.4132995745167136, "epoch": 5.75098814229249, "frac_reward_zero_std": 0.0, "grad_norm": 2.15752911567688, "learning_rate": 2e-05, "loss": 0.0003753662109375, "num_tokens": 86915428.0, "reward": 1.018890380859375, "reward/completion_length_max": 39.25, "reward/completion_length_mean": 35.6875, "reward/completion_length_min": 32.25, "reward/completion_length_p90": 37.5, "reward/completion_length_p95": 37.75, "reward/duplicate_query_rate": 0.0029296875, "reward/duplicate_ranked_plain_query_rate": 0.0029296875, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0009765625, "reward/low_length_query_rate": 0.0009765625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01953125, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.9140625, "reward/mean_r_ndcg": 0.826171875, "reward/mean_r_ndcg_cutoff": 7.0, "reward/mean_recall": 0.24072265625, "reward/mean_recall_at_10": 0.049560546875, "reward/mean_recall_at_100": 0.24072265625, "reward/mean_retrieval_combined_bounded": 0.02327728271484375, "reward/mean_retrieval_combined_raw": 0.031011104583740234, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1454.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0009765625, "reward/too_short_query_rate": 0.0, "reward_std": 0.2770073264837265, "rewards/retrieval_reward/mean": 1.01953125, "rewards/retrieval_reward/std": 0.27734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.520427703857422, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1501095462590456, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.497395277023315, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2469327189028263, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1455, "step_time": 14.057684528757818 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.08984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.4, "completions/min_terminated_length": 0.0, "entropy": 0.43825176954269407, "epoch": 5.770750988142293, "frac_reward_zero_std": 0.0, "grad_norm": 1.906356930732727, "learning_rate": 2e-05, "loss": 0.00059814453125, "num_tokens": 87218935.0, "reward": 1.022210693359375, "reward/completion_length_max": 39.6, "reward/completion_length_mean": 36.15, "reward/completion_length_min": 32.4, "reward/completion_length_p90": 37.6, "reward/completion_length_p95": 38.2, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.00859375, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.9671875, "reward/mean_mrr": 0.99140625, "reward/mean_ndcg_at_10": 0.91328125, "reward/mean_r_ndcg": 0.77734375, "reward/mean_r_ndcg_cutoff": 6.73125, "reward/mean_recall": 0.2427734375, "reward/mean_recall_at_10": 0.048779296875, "reward/mean_recall_at_100": 0.2427734375, "reward/mean_retrieval_combined_bounded": 0.03173828125, "reward/mean_retrieval_combined_raw": 0.04235639572143555, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99140625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99140625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1456.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.2869461953639984, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.2875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6030528068542482, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15834670513868332, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.857915306091309, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24219209849834442, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1460, "step_time": 14.048068047431297 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 42.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.76875, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.8, "completions/min_terminated_length": 0.0, "entropy": 0.4041156083345413, "epoch": 5.790513833992095, "frac_reward_zero_std": 0.025, "grad_norm": 1.853100299835205, "learning_rate": 2e-05, "loss": 0.00274658203125, "num_tokens": 87523695.0, "reward": 0.9769927978515625, "reward/completion_length_max": 42.2, "reward/completion_length_mean": 36.8, "reward/completion_length_min": 32.8, "reward/completion_length_p90": 39.6, "reward/completion_length_p95": 40.0, "reward/duplicate_query_rate": 0.01640625, "reward/duplicate_ranked_plain_query_rate": 0.01640625, "reward/invalid_output_rate": 0.059375, "reward/length_out_of_range_rate": 0.04296875, "reward/low_length_query_rate": 0.04296875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.978125, "reward/mean_colbert": 0.91796875, "reward/mean_mrr": 0.93984375, "reward/mean_ndcg_at_10": 0.8359375, "reward/mean_r_ndcg": 0.6953125, "reward/mean_r_ndcg_cutoff": 5.275, "reward/mean_recall": 0.22421875, "reward/mean_recall_at_10": 0.045751953125, "reward/mean_recall_at_100": 0.22421875, "reward/mean_retrieval_combined_bounded": 0.05771484375, "reward/mean_retrieval_combined_raw": 0.111865234375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.940625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.940625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1462.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.04296875, "reward/too_short_query_rate": 0.0, "reward_std": 0.39310854077339175, "rewards/retrieval_reward/mean": 0.978125, "rewards/retrieval_reward/std": 0.393359375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5895634651184083, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1966614156961441, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.468011951446533, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2335108369588852, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1465, "step_time": 14.128557215421461 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 35.4390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.2, "completions/min_terminated_length": 0.0, "entropy": 0.4215138144791126, "epoch": 5.810276679841897, "frac_reward_zero_std": 0.0, "grad_norm": 1.7858513593673706, "learning_rate": 2e-05, "loss": 0.0018798828125, "num_tokens": 87826465.0, "reward": 1.0253875732421875, "reward/completion_length_max": 40.8, "reward/completion_length_mean": 35.4, "reward/completion_length_min": 32.2, "reward/completion_length_p90": 38.0, "reward/completion_length_p95": 38.4, "reward/duplicate_query_rate": 0.009375, "reward/duplicate_ranked_plain_query_rate": 0.009375, "reward/invalid_output_rate": 0.01328125, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.96328125, "reward/mean_mrr": 0.98671875, "reward/mean_ndcg_at_10": 0.8578125, "reward/mean_r_ndcg": 0.765625, "reward/mean_r_ndcg_cutoff": 5.63125, "reward/mean_recall": 0.2474609375, "reward/mean_recall_at_10": 0.048291015625, "reward/mean_recall_at_100": 0.2474609375, "reward/mean_retrieval_combined_bounded": 0.04058837890625, "reward/mean_retrieval_combined_raw": 0.06450143456459045, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98671875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98671875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1467.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00390625, "reward/too_short_query_rate": 0.0, "reward_std": 0.2965393602848053, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8127782344818115, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1676686227321625, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.684388732910156, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2484985262155533, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1470, "step_time": 14.020284274383448 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.2, "completions/min_terminated_length": 0.0, "entropy": 0.38070125579833985, "epoch": 5.830039525691699, "frac_reward_zero_std": 0.0, "grad_norm": 1.888084053993225, "learning_rate": 2e-05, "loss": 0.00267333984375, "num_tokens": 88127653.0, "reward": 1.0258544921875, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 34.3, "reward/completion_length_min": 31.2, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 37.2, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.8921875, "reward/mean_r_ndcg": 0.77578125, "reward/mean_r_ndcg_cutoff": 6.53125, "reward/mean_recall": 0.232421875, "reward/mean_recall_at_10": 0.047900390625, "reward/mean_recall_at_100": 0.232421875, "reward/mean_retrieval_combined_bounded": 0.0337646484375, "reward/mean_retrieval_combined_raw": 0.04418937899172306, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1472.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.28470146656036377, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.28515625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8560115337371825, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20287100672721864, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.048833656311036, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23506045043468476, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1475, "step_time": 13.982244579005055 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.7328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.8, "completions/min_terminated_length": 0.0, "entropy": 0.39810906499624255, "epoch": 5.849802371541502, "frac_reward_zero_std": 0.0, "grad_norm": 1.7943546772003174, "learning_rate": 2e-05, "loss": -0.00125732421875, "num_tokens": 88427119.0, "reward": 1.015496826171875, "reward/completion_length_max": 36.4, "reward/completion_length_mean": 32.75, "reward/completion_length_min": 29.8, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.878125, "reward/mean_r_ndcg": 0.76875, "reward/mean_r_ndcg_cutoff": 6.15, "reward/mean_recall": 0.2359375, "reward/mean_recall_at_10": 0.048193359375, "reward/mean_recall_at_100": 0.2359375, "reward/mean_retrieval_combined_bounded": 0.019329833984375, "reward/mean_retrieval_combined_raw": 0.023339025676250458, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1476.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.279814088344574, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.280078125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7151753425598146, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.19505192339420319, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.700281715393066, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23790462017059327, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1480, "step_time": 13.831719236413482 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.6359375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.2, "completions/min_terminated_length": 0.0, "entropy": 0.39072548002004626, "epoch": 5.869565217391305, "frac_reward_zero_std": 0.0, "grad_norm": 2.183415651321411, "learning_rate": 2e-05, "loss": -0.000640869140625, "num_tokens": 88726461.0, "reward": 1.0189971923828125, "reward/completion_length_max": 36.2, "reward/completion_length_mean": 32.6, "reward/completion_length_min": 29.2, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01796875, "reward/mean_colbert": 0.975, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.89453125, "reward/mean_r_ndcg": 0.778125, "reward/mean_r_ndcg_cutoff": 5.8375, "reward/mean_recall": 0.23125, "reward/mean_recall_at_10": 0.048291015625, "reward/mean_recall_at_100": 0.23125, "reward/mean_retrieval_combined_bounded": 0.0227294921875, "reward/mean_retrieval_combined_raw": -0.009278273582458496, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1481.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28874475359916685, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.2890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.631992292404175, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1698673129081726, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 16.287970542907715, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2641445279121399, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1485, "step_time": 13.801923349371645 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.0859375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.2, "completions/min_terminated_length": 0.0, "entropy": 0.40052206367254256, "epoch": 5.8893280632411065, "frac_reward_zero_std": 0.0, "grad_norm": 1.9457145929336548, "learning_rate": 2e-05, "loss": 0.00023345947265625, "num_tokens": 89025163.0, "reward": 1.0155975341796875, "reward/completion_length_max": 36.0, "reward/completion_length_mean": 32.1, "reward/completion_length_min": 29.2, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 34.6, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.9, "reward/mean_r_ndcg": 0.7796875, "reward/mean_r_ndcg_cutoff": 5.340625, "reward/mean_recall": 0.22734375, "reward/mean_recall_at_10": 0.04873046875, "reward/mean_recall_at_100": 0.22734375, "reward/mean_retrieval_combined_bounded": 0.0172607421875, "reward/mean_retrieval_combined_raw": 0.012455976009368897, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1488.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2811805307865143, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.281640625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.811391258239746, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18826523125171662, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.471690464019776, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25974873900413514, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1490, "step_time": 13.770052741037215 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.4578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.4208576761186123, "epoch": 5.909090909090909, "frac_reward_zero_std": 0.0, "grad_norm": 2.462930202484131, "learning_rate": 2e-05, "loss": 0.001177978515625, "num_tokens": 89322837.0, "reward": 1.0163909912109375, "reward/completion_length_max": 35.8, "reward/completion_length_mean": 31.45, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 33.6, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.01015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.965625, "reward/mean_mrr": 0.98984375, "reward/mean_ndcg_at_10": 0.90234375, "reward/mean_r_ndcg": 0.840625, "reward/mean_r_ndcg_cutoff": 6.525, "reward/mean_recall": 0.2349609375, "reward/mean_recall_at_10": 0.048193359375, "reward/mean_recall_at_100": 0.2349609375, "reward/mean_retrieval_combined_bounded": 0.026412487030029297, "reward/mean_retrieval_combined_raw": 0.07915044128894806, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1491.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2990281045436859, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.299609375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.592471694946289, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1791888952255249, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.577554512023926, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25959521234035493, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1495, "step_time": 13.832289800187572 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.74140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.37455044910311697, "epoch": 5.928853754940711, "frac_reward_zero_std": 0.0, "grad_norm": 1.8939436674118042, "learning_rate": 2e-05, "loss": 0.0004669189453125, "num_tokens": 89620938.0, "reward": 1.0221466064453124, "reward/completion_length_max": 36.8, "reward/completion_length_mean": 31.725, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.8859375, "reward/mean_r_ndcg": 0.7765625, "reward/mean_r_ndcg_cutoff": 5.65, "reward/mean_recall": 0.2130859375, "reward/mean_recall_at_10": 0.048779296875, "reward/mean_recall_at_100": 0.2130859375, "reward/mean_retrieval_combined_bounded": 0.02723388671875, "reward/mean_retrieval_combined_raw": 0.030102584324777127, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1496.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2750110119581223, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.275, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8417502880096435, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.17220444977283478, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.368346118927002, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2710813283920288, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1500, "step_time": 13.811610680958257 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.4088400714099407, "epoch": 5.932806324110672, "frac_reward_zero_std": 0.0, "num_tokens": 89680142.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6722880677603934, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7243639022846914, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9015817223198594, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9956063268892794, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6020363019513858, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6558792748102105, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.84, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.32461045525309723, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3660174922871074, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5373493975903615, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8530120481927711, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6229519825493254, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6317748567332384, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7681250381074324, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9028755868544601, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5764833683817528, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3594578716248037, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17996070200121345, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3219009863761002, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 3.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3449.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.4130007639563307, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.44817684803549346, "retrieval_eval/nq/single_turn/recall_at_10": 0.6440272542766019, "retrieval_eval/nq/single_turn/recall_at_100": 0.8887600270609839, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6672577277029815, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6968167469167913, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8223333333333332, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9533333333333334, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.018280029296875, "reward/completion_length_max": 35.0, "reward/completion_length_mean": 30.125, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.01953125, "reward/duplicate_ranked_plain_query_rate": 0.01953125, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.95703125, "reward/mean_mrr": 0.9765625, "reward/mean_ndcg_at_10": 0.875, "reward/mean_r_ndcg": 0.67578125, "reward/mean_r_ndcg_cutoff": 6.3125, "reward/mean_recall": 0.216796875, "reward/mean_recall_at_10": 0.046142578125, "reward/mean_recall_at_100": 0.216796875, "reward/mean_retrieval_combined_bounded": 0.037841796875, "reward/mean_retrieval_combined_raw": 0.06689453125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1504.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2787729799747467, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.279296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6187756061553955, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.11765855550765991, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.716320991516113, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2965865731239319, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1501, "step_time": 13.644049539114349 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.21484375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.75, "completions/min_terminated_length": 0.0, "entropy": 0.43220632523298264, "epoch": 5.948616600790514, "frac_reward_zero_std": 0.0, "grad_norm": 2.380931854248047, "learning_rate": 2e-05, "loss": -0.000439453125, "num_tokens": 89918058.0, "reward": 1.0237922668457031, "reward/completion_length_max": 36.0, "reward/completion_length_mean": 31.25, "reward/completion_length_min": 28.75, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 33.5, "reward/duplicate_query_rate": 0.0048828125, "reward/duplicate_ranked_plain_query_rate": 0.0048828125, "reward/invalid_output_rate": 0.0048828125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.9716796875, "reward/mean_mrr": 0.9951171875, "reward/mean_ndcg_at_10": 0.931640625, "reward/mean_r_ndcg": 0.7724609375, "reward/mean_r_ndcg_cutoff": 6.640625, "reward/mean_recall": 0.23388671875, "reward/mean_recall_at_10": 0.04864501953125, "reward/mean_recall_at_100": 0.23388671875, "reward/mean_retrieval_combined_bounded": 0.028665542602539062, "reward/mean_retrieval_combined_raw": 0.03448510356247425, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9951171875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9951171875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1504.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2902955114841461, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.29052734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8226268887519836, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16418554820120335, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.396549701690674, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2832639068365097, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1505, "step_time": 13.81243114254903 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.87421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.43707256242632864, "epoch": 5.968379446640316, "frac_reward_zero_std": 0.0, "grad_norm": 2.6116859912872314, "learning_rate": 2e-05, "loss": 0.00120849609375, "num_tokens": 90215913.0, "reward": 1.0122589111328124, "reward/completion_length_max": 37.4, "reward/completion_length_mean": 31.825, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.0125, "reward/duplicate_ranked_plain_query_rate": 0.0125, "reward/invalid_output_rate": 0.0125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0109375, "reward/mean_colbert": 0.9640625, "reward/mean_mrr": 0.9875, "reward/mean_ndcg_at_10": 0.8671875, "reward/mean_r_ndcg": 0.70703125, "reward/mean_r_ndcg_cutoff": 5.6125, "reward/mean_recall": 0.215625, "reward/mean_recall_at_10": 0.048046875, "reward/mean_recall_at_100": 0.215625, "reward/mean_retrieval_combined_bounded": 0.0248291015625, "reward/mean_retrieval_combined_raw": 0.059814453125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1505.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2976708233356476, "rewards/retrieval_reward/mean": 1.0109375, "rewards/retrieval_reward/std": 0.2984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.711440992355347, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.16189926862716675, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.975994300842284, "sampling/raw_vllm_sampling_logp_difference/mean": 0.28089306354522703, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1510, "step_time": 13.92369701693533 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.0765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.4048777617514133, "epoch": 5.988142292490118, "frac_reward_zero_std": 0.0, "grad_norm": 2.3194737434387207, "learning_rate": 2e-05, "loss": -9.918212890625e-05, "num_tokens": 90514603.0, "reward": 1.0146240234375, "reward/completion_length_max": 36.2, "reward/completion_length_mean": 32.075, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 34.4, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.8984375, "reward/mean_r_ndcg": 0.77578125, "reward/mean_r_ndcg_cutoff": 6.49375, "reward/mean_recall": 0.22578125, "reward/mean_recall_at_10": 0.04892578125, "reward/mean_recall_at_100": 0.22578125, "reward/mean_retrieval_combined_bounded": 0.02086181640625, "reward/mean_retrieval_combined_raw": 0.004296875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1512.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28674508929252623, "rewards/retrieval_reward/mean": 1.01328125, "rewards/retrieval_reward/std": 0.287109375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4812610864639284, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13021864593029023, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.793253707885743, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26972718834877013, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1515, "step_time": 13.907315546402241 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.4703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.8, "completions/min_terminated_length": 0.0, "entropy": 0.3829656526446342, "epoch": 6.007905138339921, "frac_reward_zero_std": 0.0, "grad_norm": 2.072399854660034, "learning_rate": 2e-05, "loss": 0.00027008056640625, "num_tokens": 90812453.0, "reward": 1.028851318359375, "reward/completion_length_max": 36.8, "reward/completion_length_mean": 31.475, "reward/completion_length_min": 28.8, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.0125, "reward/duplicate_ranked_plain_query_rate": 0.0125, "reward/invalid_output_rate": 0.0125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.028125, "reward/mean_colbert": 0.9625, "reward/mean_mrr": 0.9875, "reward/mean_ndcg_at_10": 0.86640625, "reward/mean_r_ndcg": 0.70078125, "reward/mean_r_ndcg_cutoff": 5.45625, "reward/mean_recall": 0.244140625, "reward/mean_recall_at_10": 0.048291015625, "reward/mean_recall_at_100": 0.244140625, "reward/mean_retrieval_combined_bounded": 0.041259765625, "reward/mean_retrieval_combined_raw": 0.05625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1518.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28495020270347593, "rewards/retrieval_reward/mean": 1.028125, "rewards/retrieval_reward/std": 0.285546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.649692249298096, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1768186330795288, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.491240692138671, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27049326300621035, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1520, "step_time": 14.048798922554123 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.16953125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.37766825035214424, "epoch": 6.027667984189724, "frac_reward_zero_std": 0.0, "grad_norm": 2.243152618408203, "learning_rate": 2e-05, "loss": -0.00018768310546875, "num_tokens": 91110302.0, "reward": 1.01998291015625, "reward/completion_length_max": 37.4, "reward/completion_length_mean": 31.175, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 33.4, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.975, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.90859375, "reward/mean_r_ndcg": 0.8125, "reward/mean_r_ndcg_cutoff": 6.2625, "reward/mean_recall": 0.215625, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.215625, "reward/mean_retrieval_combined_bounded": 0.0214111328125, "reward/mean_retrieval_combined_raw": 0.008858457952737809, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1520.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.26612708270549773, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.2662109375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.524764966964722, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18423212468624114, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.098619079589843, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2350892335176468, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1525, "step_time": 14.021183678426315 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.18046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.3836915299296379, "epoch": 6.047430830039525, "frac_reward_zero_std": 0.0, "grad_norm": 2.5538532733917236, "learning_rate": 2e-05, "loss": 0.000970458984375, "num_tokens": 91407845.0, "reward": 1.007867431640625, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 31.225, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 33.4, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0078125, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.99765625, "reward/mean_ndcg_at_10": 0.8765625, "reward/mean_r_ndcg": 0.73984375, "reward/mean_r_ndcg_cutoff": 5.6875, "reward/mean_recall": 0.23515625, "reward/mean_recall_at_10": 0.04853515625, "reward/mean_recall_at_100": 0.23515625, "reward/mean_retrieval_combined_bounded": 0.0102691650390625, "reward/mean_retrieval_combined_raw": 0.015229468513280154, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1528.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2841761648654938, "rewards/retrieval_reward/mean": 1.0078125, "rewards/retrieval_reward/std": 0.284765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8601471424102782, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.23071117997169494, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.74266357421875, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21382997334003448, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1530, "step_time": 14.043882160552311 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.21796875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.8, "completions/min_terminated_length": 0.0, "entropy": 0.40799615457654, "epoch": 6.067193675889328, "frac_reward_zero_std": 0.0, "grad_norm": 2.013120412826538, "learning_rate": 2e-05, "loss": -0.00045166015625, "num_tokens": 91706812.0, "reward": 1.012335205078125, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 32.175, "reward/completion_length_min": 28.8, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.8, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.00625, "reward/low_length_query_rate": 0.00625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01328125, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.88984375, "reward/mean_r_ndcg": 0.7578125, "reward/mean_r_ndcg_cutoff": 5.6625, "reward/mean_recall": 0.2021484375, "reward/mean_recall_at_10": 0.047998046875, "reward/mean_recall_at_100": 0.2021484375, "reward/mean_retrieval_combined_bounded": 0.023291015625, "reward/mean_retrieval_combined_raw": 0.024810728430747987, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1532.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3031842648983002, "rewards/retrieval_reward/mean": 1.01171875, "rewards/retrieval_reward/std": 0.303515625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5094277381896974, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18764874935150147, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.680491638183593, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24909383952617645, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1535, "step_time": 14.063851232570595 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.803125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.8, "completions/min_terminated_length": 0.0, "entropy": 0.40806510373950006, "epoch": 6.086956521739131, "frac_reward_zero_std": 0.0, "grad_norm": 2.238955020904541, "learning_rate": 2e-05, "loss": -0.00027008056640625, "num_tokens": 92006400.0, "reward": 1.0183868408203125, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 32.775, "reward/completion_length_min": 29.8, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.02265625, "reward/duplicate_ranked_plain_query_rate": 0.02265625, "reward/invalid_output_rate": 0.02265625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.95390625, "reward/mean_mrr": 0.97734375, "reward/mean_ndcg_at_10": 0.9109375, "reward/mean_r_ndcg": 0.80078125, "reward/mean_r_ndcg_cutoff": 6.01875, "reward/mean_recall": 0.2251953125, "reward/mean_recall_at_10": 0.0484375, "reward/mean_recall_at_100": 0.2251953125, "reward/mean_retrieval_combined_bounded": 0.0410888671875, "reward/mean_retrieval_combined_raw": 0.07182393074035645, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97734375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97734375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1536.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3069713056087494, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.307421875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.804832410812378, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.22777141332626344, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.704974174499512, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2129715710878372, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1540, "step_time": 13.97682048495626 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.61015625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.4091191656887531, "epoch": 6.106719367588933, "frac_reward_zero_std": 0.0, "grad_norm": 2.3164010047912598, "learning_rate": 2e-05, "loss": -0.0028564453125, "num_tokens": 92306541.0, "reward": 1.033807373046875, "reward/completion_length_max": 38.2, "reward/completion_length_mean": 33.6, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 36.6, "reward/duplicate_query_rate": 0.0140625, "reward/duplicate_ranked_plain_query_rate": 0.0140625, "reward/invalid_output_rate": 0.01484375, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.034375, "reward/mean_colbert": 0.96015625, "reward/mean_mrr": 0.98515625, "reward/mean_ndcg_at_10": 0.884375, "reward/mean_r_ndcg": 0.7859375, "reward/mean_r_ndcg_cutoff": 5.69375, "reward/mean_recall": 0.2482421875, "reward/mean_recall_at_10": 0.048486328125, "reward/mean_recall_at_100": 0.2482421875, "reward/mean_retrieval_combined_bounded": 0.0490234375, "reward/mean_retrieval_combined_raw": 0.0709153175354004, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98515625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98515625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1542.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.295157390832901, "rewards/retrieval_reward/mean": 1.034375, "rewards/retrieval_reward/std": 0.295703125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7950930118560793, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21709004640579224, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.442274475097657, "sampling/raw_vllm_sampling_logp_difference/mean": 0.224896439909935, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1545, "step_time": 13.98500351877883 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.99921875, "completions/max_length": 38.6, "completions/max_terminated_length": 6.8, "completions/mean_length": 33.684375, "completions/mean_terminated_length": 6.8, "completions/min_length": 30.0, "completions/min_terminated_length": 6.8, "entropy": 0.4135304197669029, "epoch": 6.126482213438735, "frac_reward_zero_std": 0.0, "grad_norm": 2.089745044708252, "learning_rate": 2e-05, "loss": 0.00107421875, "num_tokens": 92607033.0, "reward": 1.0315582275390625, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 33.7, "reward/completion_length_min": 30.0, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.0359375, "reward/duplicate_ranked_plain_query_rate": 0.0359375, "reward/invalid_output_rate": 0.0359375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0296875, "reward/mean_colbert": 0.93984375, "reward/mean_mrr": 0.9640625, "reward/mean_ndcg_at_10": 0.828125, "reward/mean_r_ndcg": 0.6859375, "reward/mean_r_ndcg_cutoff": 5.25625, "reward/mean_recall": 0.23984375, "reward/mean_recall_at_10": 0.046875, "reward/mean_recall_at_100": 0.23984375, "reward/mean_retrieval_combined_bounded": 0.067333984375, "reward/mean_retrieval_combined_raw": 0.13046875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9640625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9640625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1547.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3166710674762726, "rewards/retrieval_reward/mean": 1.0296875, "rewards/retrieval_reward/std": 0.316796875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5110738039016725, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1886062279343605, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.710871505737305, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2304929107427597, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1550, "step_time": 14.062583317805547 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 42.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.9375, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.0, "completions/min_terminated_length": 0.0, "entropy": 0.38966719806194305, "epoch": 6.130434782608695, "frac_reward_zero_std": 0.0, "num_tokens": 92667561.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6685042433544295, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7224512456495804, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9050966608084359, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9973637961335676, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6218709518975012, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6678662032755966, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8333333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.32724086303684247, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.36256427749526277, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5180722891566265, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8578313253012049, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6386050310669893, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6408634994333663, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7643790774952747, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9158702213279677, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5795125053436311, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.36369916909185207, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17515778042163738, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3185157423175446, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 4.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3448.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.4206434442764117, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.45828272579859874, "retrieval_eval/nq/single_turn/recall_at_10": 0.6586910286156226, "retrieval_eval/nq/single_turn/recall_at_100": 0.896606728538283, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6771753522976199, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7089560174045885, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8456666666666667, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9626666666666667, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.01422119140625, "reward/completion_length_max": 42.0, "reward/completion_length_mean": 35.0, "reward/completion_length_min": 32.0, "reward/completion_length_p90": 37.0, "reward/completion_length_p95": 38.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.01953125, "reward/length_out_of_range_rate": 0.015625, "reward/low_length_query_rate": 0.015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.95703125, "reward/mean_mrr": 0.98046875, "reward/mean_ndcg_at_10": 0.8515625, "reward/mean_r_ndcg": 0.7421875, "reward/mean_r_ndcg_cutoff": 5.625, "reward/mean_recall": 0.2490234375, "reward/mean_recall_at_10": 0.049072265625, "reward/mean_recall_at_100": 0.2490234375, "reward/mean_retrieval_combined_bounded": 0.041748046875, "reward/mean_retrieval_combined_raw": 0.11328125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1552.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3119686543941498, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.3125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6865310668945312, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20593217015266418, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.017467498779297, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23158162832260132, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1551, "step_time": 14.521678477001842 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.3349609375, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.25, "completions/min_terminated_length": 0.0, "entropy": 0.4025421803817153, "epoch": 6.146245059288537, "frac_reward_zero_std": 0.0, "grad_norm": 2.1270251274108887, "learning_rate": 2e-05, "loss": -0.001123046875, "num_tokens": 92907712.0, "reward": 1.0297966003417969, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 33.4375, "reward/completion_length_min": 30.25, "reward/completion_length_p90": 35.5, "reward/completion_length_p95": 36.25, "reward/duplicate_query_rate": 0.0166015625, "reward/duplicate_ranked_plain_query_rate": 0.0166015625, "reward/invalid_output_rate": 0.0166015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.029296875, "reward/mean_colbert": 0.958984375, "reward/mean_mrr": 0.9833984375, "reward/mean_ndcg_at_10": 0.88671875, "reward/mean_r_ndcg": 0.7646484375, "reward/mean_r_ndcg_cutoff": 5.9375, "reward/mean_recall": 0.252685546875, "reward/mean_recall_at_10": 0.04852294921875, "reward/mean_recall_at_100": 0.252685546875, "reward/mean_retrieval_combined_bounded": 0.0464019775390625, "reward/mean_retrieval_combined_raw": 0.07733154296875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9833984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9833984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1552.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2891940474510193, "rewards/retrieval_reward/mean": 1.029296875, "rewards/retrieval_reward/std": 0.2900390625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7689096927642822, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20289908349514008, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.759863257408142, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22495358064770699, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1555, "step_time": 14.01321811901289 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.81328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.8, "completions/min_terminated_length": 0.0, "entropy": 0.39321795105934143, "epoch": 6.16600790513834, "frac_reward_zero_std": 0.0, "grad_norm": 1.7291078567504883, "learning_rate": 2e-05, "loss": 0.00179443359375, "num_tokens": 93208273.0, "reward": 0.996734619140625, "reward/completion_length_max": 39.6, "reward/completion_length_mean": 33.8, "reward/completion_length_min": 29.8, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 37.2, "reward/duplicate_query_rate": 0.03671875, "reward/duplicate_ranked_plain_query_rate": 0.03671875, "reward/invalid_output_rate": 0.04375, "reward/length_out_of_range_rate": 0.00703125, "reward/low_length_query_rate": 0.00703125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.996875, "reward/mean_colbert": 0.93203125, "reward/mean_mrr": 0.95625, "reward/mean_ndcg_at_10": 0.8609375, "reward/mean_r_ndcg": 0.69765625, "reward/mean_r_ndcg_cutoff": 5.45625, "reward/mean_recall": 0.24375, "reward/mean_recall_at_10": 0.04736328125, "reward/mean_recall_at_100": 0.24375, "reward/mean_retrieval_combined_bounded": 0.043994140625, "reward/mean_retrieval_combined_raw": 0.091943359375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.95625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.95625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1556.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00703125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3424277424812317, "rewards/retrieval_reward/mean": 0.996875, "rewards/retrieval_reward/std": 0.343359375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7427664279937742, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21415568590164186, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.055629348754882, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2220038056373596, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1560, "step_time": 13.986209477006923 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 57.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.09609375, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.4, "completions/min_terminated_length": 0.0, "entropy": 0.39880235195159913, "epoch": 6.1857707509881426, "frac_reward_zero_std": 0.0, "grad_norm": 2.179853677749634, "learning_rate": 2e-05, "loss": 0.02099609375, "num_tokens": 93509740.0, "reward": 1.017791748046875, "reward/completion_length_max": 57.4, "reward/completion_length_mean": 34.05, "reward/completion_length_min": 30.4, "reward/completion_length_p90": 37.6, "reward/completion_length_p95": 37.6, "reward/duplicate_query_rate": 0.02578125, "reward/duplicate_ranked_plain_query_rate": 0.02578125, "reward/invalid_output_rate": 0.03125, "reward/length_out_of_range_rate": 0.003125, "reward/low_length_query_rate": 0.003125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01796875, "reward/mean_colbert": 0.946875, "reward/mean_mrr": 0.96875, "reward/mean_ndcg_at_10": 0.8921875, "reward/mean_r_ndcg": 0.75078125, "reward/mean_r_ndcg_cutoff": 6.30625, "reward/mean_recall": 0.2130859375, "reward/mean_recall_at_10": 0.04794921875, "reward/mean_recall_at_100": 0.2130859375, "reward/mean_retrieval_combined_bounded": 0.055224609375, "reward/mean_retrieval_combined_raw": 0.09077112525701522, "reward/output_tag_rate": 0.99765625, "reward/query_valid_rate": 0.96875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1561.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.003125, "reward/too_short_query_rate": 0.0, "reward_std": 0.35250550508499146, "rewards/retrieval_reward/mean": 1.01796875, "rewards/retrieval_reward/std": 0.353125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7802924633026125, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.22295311689376832, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.438722229003906, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2211791306734085, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1565, "step_time": 15.04344527766807 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.62890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.40986790135502815, "epoch": 6.205533596837944, "frac_reward_zero_std": 0.0, "grad_norm": 1.9988112449645996, "learning_rate": 2e-05, "loss": -0.000604248046875, "num_tokens": 93810161.0, "reward": 1.0133758544921876, "reward/completion_length_max": 39.6, "reward/completion_length_mean": 33.6, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 36.4, "reward/completion_length_p95": 36.8, "reward/duplicate_query_rate": 0.0234375, "reward/duplicate_ranked_plain_query_rate": 0.0234375, "reward/invalid_output_rate": 0.0234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.9515625, "reward/mean_mrr": 0.9765625, "reward/mean_ndcg_at_10": 0.903125, "reward/mean_r_ndcg": 0.78203125, "reward/mean_r_ndcg_cutoff": 6.35, "reward/mean_recall": 0.2361328125, "reward/mean_recall_at_10": 0.047509765625, "reward/mean_recall_at_100": 0.2361328125, "reward/mean_retrieval_combined_bounded": 0.03681640625, "reward/mean_retrieval_combined_raw": 0.07770996801555156, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1568.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.30726192593574525, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.3072265625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5873124599456787, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18073202669620514, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.232010650634766, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23159097135066986, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1570, "step_time": 14.193919016234577 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.8609375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.8, "completions/min_terminated_length": 0.0, "entropy": 0.4219138391315937, "epoch": 6.225296442687747, "frac_reward_zero_std": 0.0, "grad_norm": 1.7734012603759766, "learning_rate": 2e-05, "loss": 0.00029296875, "num_tokens": 94111135.0, "reward": 1.031494140625, "reward/completion_length_max": 39.0, "reward/completion_length_mean": 33.85, "reward/completion_length_min": 29.8, "reward/completion_length_p90": 36.6, "reward/completion_length_p95": 37.2, "reward/duplicate_query_rate": 0.0265625, "reward/duplicate_ranked_plain_query_rate": 0.0265625, "reward/invalid_output_rate": 0.0265625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0328125, "reward/mean_colbert": 0.9515625, "reward/mean_mrr": 0.9734375, "reward/mean_ndcg_at_10": 0.86875, "reward/mean_r_ndcg": 0.76171875, "reward/mean_r_ndcg_cutoff": 5.83125, "reward/mean_recall": 0.2259765625, "reward/mean_recall_at_10": 0.04775390625, "reward/mean_recall_at_100": 0.2259765625, "reward/mean_retrieval_combined_bounded": 0.0580322265625, "reward/mean_retrieval_combined_raw": 0.11534395217895507, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9734375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9734375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1571.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.30480563044548037, "rewards/retrieval_reward/mean": 1.0328125, "rewards/retrieval_reward/std": 0.30546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7144152164459228, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13771027624607085, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.987504196166991, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26938636302948, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1575, "step_time": 14.087087708979379 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.365625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.0, "completions/min_terminated_length": 0.0, "entropy": 0.3999443888664246, "epoch": 6.24505928853755, "frac_reward_zero_std": 0.0, "grad_norm": 1.867202877998352, "learning_rate": 2e-05, "loss": 0.0005401611328125, "num_tokens": 94411379.0, "reward": 1.013482666015625, "reward/completion_length_max": 37.6, "reward/completion_length_mean": 33.35, "reward/completion_length_min": 30.0, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.01015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01171875, "reward/mean_colbert": 0.965625, "reward/mean_mrr": 0.98984375, "reward/mean_ndcg_at_10": 0.890625, "reward/mean_r_ndcg": 0.77890625, "reward/mean_r_ndcg_cutoff": 6.5125, "reward/mean_recall": 0.2388671875, "reward/mean_recall_at_10": 0.047998046875, "reward/mean_recall_at_100": 0.2388671875, "reward/mean_retrieval_combined_bounded": 0.0236785888671875, "reward/mean_retrieval_combined_raw": 0.029528382420539855, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1576.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2892824649810791, "rewards/retrieval_reward/mean": 1.01328125, "rewards/retrieval_reward/std": 0.28984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5503652572631834, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1783807635307312, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.437299728393555, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2517568141222, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1580, "step_time": 14.001751888205764 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.325, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.4, "completions/min_terminated_length": 0.0, "entropy": 0.3884332440793514, "epoch": 6.264822134387352, "frac_reward_zero_std": 0.0, "grad_norm": 1.8631961345672607, "learning_rate": 2e-05, "loss": 0.00035858154296875, "num_tokens": 94711475.0, "reward": 1.0404388427734375, "reward/completion_length_max": 38.6, "reward/completion_length_mean": 33.3, "reward/completion_length_min": 30.4, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.02421875, "reward/duplicate_ranked_plain_query_rate": 0.02421875, "reward/invalid_output_rate": 0.02734375, "reward/length_out_of_range_rate": 0.003125, "reward/low_length_query_rate": 0.003125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.040625, "reward/mean_colbert": 0.94921875, "reward/mean_mrr": 0.97265625, "reward/mean_ndcg_at_10": 0.89609375, "reward/mean_r_ndcg": 0.7453125, "reward/mean_r_ndcg_cutoff": 6.36875, "reward/mean_recall": 0.20390625, "reward/mean_recall_at_10": 0.047802734375, "reward/mean_recall_at_100": 0.20390625, "reward/mean_retrieval_combined_bounded": 0.06943359375, "reward/mean_retrieval_combined_raw": 0.19001197814941406, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97265625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97265625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1584.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.003125, "reward/too_short_query_rate": 0.0, "reward_std": 0.3285678267478943, "rewards/retrieval_reward/mean": 1.040625, "rewards/retrieval_reward/std": 0.32890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8889538764953615, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1744722992181778, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.334160423278808, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24337629079818726, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1585, "step_time": 13.962714034749661 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.65234375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.8, "completions/min_terminated_length": 0.0, "entropy": 0.36481920406222346, "epoch": 6.284584980237154, "frac_reward_zero_std": 0.0, "grad_norm": 2.3747105598449707, "learning_rate": 2e-05, "loss": 0.001263427734375, "num_tokens": 95010614.0, "reward": 1.040936279296875, "reward/completion_length_max": 38.2, "reward/completion_length_mean": 32.65, "reward/completion_length_min": 29.8, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.0171875, "reward/duplicate_ranked_plain_query_rate": 0.0171875, "reward/invalid_output_rate": 0.0171875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.040625, "reward/mean_colbert": 0.959375, "reward/mean_mrr": 0.9828125, "reward/mean_ndcg_at_10": 0.8921875, "reward/mean_r_ndcg": 0.7859375, "reward/mean_r_ndcg_cutoff": 6.6625, "reward/mean_recall": 0.2283203125, "reward/mean_recall_at_10": 0.04716796875, "reward/mean_recall_at_100": 0.2283203125, "reward/mean_retrieval_combined_bounded": 0.0580810546875, "reward/mean_retrieval_combined_raw": 0.109716796875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1585.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3037609398365021, "rewards/retrieval_reward/mean": 1.040625, "rewards/retrieval_reward/std": 0.304296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.687636613845825, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.17558335065841674, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.705896759033203, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24280603528022765, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1590, "step_time": 14.093035106756725 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.675, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.6, "completions/min_terminated_length": 0.0, "entropy": 0.3727143481373787, "epoch": 6.304347826086957, "frac_reward_zero_std": 0.0, "grad_norm": 2.2929670810699463, "learning_rate": 2e-05, "loss": -0.00019073486328125, "num_tokens": 95309942.0, "reward": 1.016204833984375, "reward/completion_length_max": 37.6, "reward/completion_length_mean": 32.7, "reward/completion_length_min": 29.6, "reward/completion_length_p90": 35.4, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.00859375, "reward/duplicate_ranked_plain_query_rate": 0.00859375, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.96640625, "reward/mean_mrr": 0.98203125, "reward/mean_ndcg_at_10": 0.82109375, "reward/mean_r_ndcg": 0.67734375, "reward/mean_r_ndcg_cutoff": 5.30625, "reward/mean_recall": 0.2251953125, "reward/mean_recall_at_10": 0.047412109375, "reward/mean_recall_at_100": 0.2251953125, "reward/mean_retrieval_combined_bounded": 0.02608642578125, "reward/mean_retrieval_combined_raw": 0.04483342170715332, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1592.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.28494860529899596, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.285546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.628691005706787, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.19457471370697021, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.032976150512695, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24958646595478057, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1595, "step_time": 13.957359555223956 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.28046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.6, "completions/min_terminated_length": 0.0, "entropy": 0.3522986602038145, "epoch": 6.324110671936759, "frac_reward_zero_std": 0.0, "grad_norm": 2.0735132694244385, "learning_rate": 2e-05, "loss": 0.000799560546875, "num_tokens": 95609629.0, "reward": 1.0198883056640624, "reward/completion_length_max": 39.4, "reward/completion_length_mean": 33.3, "reward/completion_length_min": 29.6, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.8, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.8890625, "reward/mean_r_ndcg": 0.771875, "reward/mean_r_ndcg_cutoff": 6.38125, "reward/mean_recall": 0.2384765625, "reward/mean_recall_at_10": 0.048291015625, "reward/mean_recall_at_100": 0.2384765625, "reward/mean_retrieval_combined_bounded": 0.029638671875, "reward/mean_retrieval_combined_raw": 0.04243628978729248, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1598.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00390625, "reward/too_short_query_rate": 0.0, "reward_std": 0.27766804695129393, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.2783203125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8257468700408936, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2360999256372452, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.279302406311036, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21577182710170745, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1600, "step_time": 14.070816216338425 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.45703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.0, "completions/min_terminated_length": 0.0, "entropy": 0.3522401228547096, "epoch": 6.32806324110672, "frac_reward_zero_std": 0.0, "num_tokens": 95669650.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6706862705971998, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7234310291217115, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9024604569420035, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9964850615114236, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5895277816743582, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.645752991383426, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.84, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 1.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.34347020586311094, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3771195570283688, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5253012048192771, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8530120481927711, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6241814466974765, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6281107997641929, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.742770257911103, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9080734406438633, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5655937515703592, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3533569344822806, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17235930355573986, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3096918498194637, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 1.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3451.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.4269480636517067, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4628588330657245, "retrieval_eval/nq/single_turn/recall_at_10": 0.657707910750507, "retrieval_eval/nq/single_turn/recall_at_100": 0.8874963778614894, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6790157474947861, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7083622273235612, "retrieval_eval/scifact/single_turn/recall_at_10": 0.839, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9376666666666668, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0293121337890625, "reward/completion_length_max": 40.0, "reward/completion_length_mean": 33.5, "reward/completion_length_min": 30.0, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.88671875, "reward/mean_r_ndcg": 0.84765625, "reward/mean_r_ndcg_cutoff": 6.9375, "reward/mean_recall": 0.26171875, "reward/mean_recall_at_10": 0.04931640625, "reward/mean_recall_at_100": 0.26171875, "reward/mean_retrieval_combined_bounded": 0.03759765625, "reward/mean_retrieval_combined_raw": 0.058837890625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1600.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2715509533882141, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.271484375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.9884328842163086, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.23173609375953674, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.315762519836426, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20789451897144318, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1601, "step_time": 14.133025304006878 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.5, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.3671875, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.25, "completions/min_terminated_length": 0.0, "entropy": 0.360470580868423, "epoch": 6.3438735177865615, "frac_reward_zero_std": 0.0, "grad_norm": 2.1961777210235596, "learning_rate": 2e-05, "loss": 0.00084228515625, "num_tokens": 95910922.0, "reward": 1.0250167846679688, "reward/completion_length_max": 41.5, "reward/completion_length_mean": 34.375, "reward/completion_length_min": 30.25, "reward/completion_length_p90": 36.75, "reward/completion_length_p95": 37.75, "reward/duplicate_query_rate": 0.0166015625, "reward/duplicate_ranked_plain_query_rate": 0.0166015625, "reward/invalid_output_rate": 0.021484375, "reward/length_out_of_range_rate": 0.0048828125, "reward/low_length_query_rate": 0.0048828125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025390625, "reward/mean_colbert": 0.9541015625, "reward/mean_mrr": 0.978515625, "reward/mean_ndcg_at_10": 0.9189453125, "reward/mean_r_ndcg": 0.7978515625, "reward/mean_r_ndcg_cutoff": 6.3125, "reward/mean_recall": 0.22412109375, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.22412109375, "reward/mean_retrieval_combined_bounded": 0.0489349365234375, "reward/mean_retrieval_combined_raw": 0.0882578045129776, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.978515625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.978515625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1600.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0048828125, "reward/too_short_query_rate": 0.0, "reward_std": 0.31675921380519867, "rewards/retrieval_reward/mean": 1.025390625, "rewards/retrieval_reward/std": 0.3173828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.794638693332672, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18281245604157448, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.373536109924316, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23493816331028938, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1605, "step_time": 14.17224049649667 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.53046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.2, "completions/min_terminated_length": 0.0, "entropy": 0.4167686626315117, "epoch": 6.363636363636363, "frac_reward_zero_std": 0.0, "grad_norm": 2.211439371109009, "learning_rate": 2e-05, "loss": 0.001806640625, "num_tokens": 96212849.0, "reward": 1.0295501708984376, "reward/completion_length_max": 40.6, "reward/completion_length_mean": 34.55, "reward/completion_length_min": 31.2, "reward/completion_length_p90": 36.8, "reward/completion_length_p95": 37.6, "reward/duplicate_query_rate": 0.02109375, "reward/duplicate_ranked_plain_query_rate": 0.02109375, "reward/invalid_output_rate": 0.02734375, "reward/length_out_of_range_rate": 0.00625, "reward/low_length_query_rate": 0.00625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0296875, "reward/mean_colbert": 0.94921875, "reward/mean_mrr": 0.97265625, "reward/mean_ndcg_at_10": 0.9, "reward/mean_r_ndcg": 0.75234375, "reward/mean_r_ndcg_cutoff": 6.725, "reward/mean_recall": 0.2162109375, "reward/mean_recall_at_10": 0.047900390625, "reward/mean_recall_at_100": 0.2162109375, "reward/mean_retrieval_combined_bounded": 0.0599609375, "reward/mean_retrieval_combined_raw": 0.10947265625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97265625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97265625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1608.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3155448317527771, "rewards/retrieval_reward/mean": 1.0296875, "rewards/retrieval_reward/std": 0.316015625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.444776773452759, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12710507810115815, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.618792915344239, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2930128961801529, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1610, "step_time": 14.065061460027938 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 35.36875, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.2, "completions/min_terminated_length": 0.0, "entropy": 0.3683372184634209, "epoch": 6.383399209486166, "frac_reward_zero_std": 0.0, "grad_norm": 2.5116770267486572, "learning_rate": 2e-05, "loss": 0.00091552734375, "num_tokens": 96516073.0, "reward": 1.026348876953125, "reward/completion_length_max": 39.8, "reward/completion_length_mean": 35.35, "reward/completion_length_min": 32.2, "reward/completion_length_p90": 37.8, "reward/completion_length_p95": 38.4, "reward/duplicate_query_rate": 0.01875, "reward/duplicate_ranked_plain_query_rate": 0.01875, "reward/invalid_output_rate": 0.03046875, "reward/length_out_of_range_rate": 0.01171875, "reward/low_length_query_rate": 0.01171875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.028125, "reward/mean_colbert": 0.9453125, "reward/mean_mrr": 0.96953125, "reward/mean_ndcg_at_10": 0.87890625, "reward/mean_r_ndcg": 0.74140625, "reward/mean_r_ndcg_cutoff": 6.975, "reward/mean_recall": 0.2326171875, "reward/mean_recall_at_10": 0.047314453125, "reward/mean_recall_at_100": 0.2326171875, "reward/mean_retrieval_combined_bounded": 0.06279296875, "reward/mean_retrieval_combined_raw": 0.12041015625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.96953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1612.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.01171875, "reward/too_short_query_rate": 0.0, "reward_std": 0.3431513667106628, "rewards/retrieval_reward/mean": 1.028125, "rewards/retrieval_reward/std": 0.343359375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7728281021118164, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15159263759851455, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.587066459655762, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2628666371107101, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1615, "step_time": 14.09752450583037 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 35.31640625, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.0, "completions/min_terminated_length": 0.0, "entropy": 0.4180549085140228, "epoch": 6.403162055335969, "frac_reward_zero_std": 0.0, "grad_norm": 2.465132474899292, "learning_rate": 2e-05, "loss": 0.00118408203125, "num_tokens": 96818654.0, "reward": 1.0314910888671875, "reward/completion_length_max": 41.0, "reward/completion_length_mean": 35.3, "reward/completion_length_min": 32.0, "reward/completion_length_p90": 37.6, "reward/completion_length_p95": 38.2, "reward/duplicate_query_rate": 0.01875, "reward/duplicate_ranked_plain_query_rate": 0.01875, "reward/invalid_output_rate": 0.02109375, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.953125, "reward/mean_mrr": 0.96953125, "reward/mean_ndcg_at_10": 0.80234375, "reward/mean_r_ndcg": 0.68046875, "reward/mean_r_ndcg_cutoff": 5.55625, "reward/mean_recall": 0.2244140625, "reward/mean_recall_at_10": 0.046044921875, "reward/mean_recall_at_100": 0.2244140625, "reward/mean_retrieval_combined_bounded": 0.0537353515625, "reward/mean_retrieval_combined_raw": 0.09111328125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1616.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.30633240938186646, "rewards/retrieval_reward/mean": 1.0296875, "rewards/retrieval_reward/std": 0.30625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5996406078338623, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12153280079364777, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.92810001373291, "sampling/raw_vllm_sampling_logp_difference/mean": 0.29694571495056155, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1620, "step_time": 14.105351333040744 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 35.9796875, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.6, "completions/min_terminated_length": 0.0, "entropy": 0.46075202375650404, "epoch": 6.4229249011857705, "frac_reward_zero_std": 0.0, "grad_norm": 2.325542449951172, "learning_rate": 2e-05, "loss": 0.001483154296875, "num_tokens": 97122372.0, "reward": 0.9942230224609375, "reward/completion_length_max": 41.4, "reward/completion_length_mean": 35.95, "reward/completion_length_min": 32.6, "reward/completion_length_p90": 38.2, "reward/completion_length_p95": 39.0, "reward/duplicate_query_rate": 0.025, "reward/duplicate_ranked_plain_query_rate": 0.025, "reward/invalid_output_rate": 0.03359375, "reward/length_out_of_range_rate": 0.00859375, "reward/low_length_query_rate": 0.00859375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99375, "reward/mean_colbert": 0.940625, "reward/mean_mrr": 0.965625, "reward/mean_ndcg_at_10": 0.865625, "reward/mean_r_ndcg": 0.76484375, "reward/mean_r_ndcg_cutoff": 6.28125, "reward/mean_recall": 0.2220703125, "reward/mean_recall_at_10": 0.046923828125, "reward/mean_recall_at_100": 0.2220703125, "reward/mean_retrieval_combined_bounded": 0.032275390625, "reward/mean_retrieval_combined_raw": -0.00068359375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.96640625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96640625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1622.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00859375, "reward/too_short_query_rate": 0.0, "reward_std": 0.3458551824092865, "rewards/retrieval_reward/mean": 0.99296875, "rewards/retrieval_reward/std": 0.34609375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5595598220825195, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.11332928389310837, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.848991012573242, "sampling/raw_vllm_sampling_logp_difference/mean": 0.29251834750175476, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1625, "step_time": 14.109828888007906 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 42.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.1765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.8, "completions/min_terminated_length": 0.0, "entropy": 0.4563277378678322, "epoch": 6.442687747035573, "frac_reward_zero_std": 0.0, "grad_norm": 2.0815629959106445, "learning_rate": 2e-05, "loss": 0.006396484375, "num_tokens": 97426150.0, "reward": 1.006060791015625, "reward/completion_length_max": 42.8, "reward/completion_length_mean": 36.2, "reward/completion_length_min": 32.8, "reward/completion_length_p90": 38.6, "reward/completion_length_p95": 39.0, "reward/duplicate_query_rate": 0.02734375, "reward/duplicate_ranked_plain_query_rate": 0.02734375, "reward/invalid_output_rate": 0.0515625, "reward/length_out_of_range_rate": 0.02421875, "reward/low_length_query_rate": 0.02421875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00703125, "reward/mean_colbert": 0.925, "reward/mean_mrr": 0.9484375, "reward/mean_ndcg_at_10": 0.8484375, "reward/mean_r_ndcg": 0.7125, "reward/mean_r_ndcg_cutoff": 5.5875, "reward/mean_recall": 0.2173828125, "reward/mean_recall_at_10": 0.045849609375, "reward/mean_recall_at_100": 0.2173828125, "reward/mean_retrieval_combined_bounded": 0.0697509765625, "reward/mean_retrieval_combined_raw": 0.1369384765625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9484375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9484375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1627.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.02421875, "reward/too_short_query_rate": 0.0, "reward_std": 0.3759737730026245, "rewards/retrieval_reward/mean": 1.00703125, "rewards/retrieval_reward/std": 0.3765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.58024640083313, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.11659692674875259, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.1681058883667, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3302451968193054, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1630, "step_time": 14.138658796611708 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 41.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.484375, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.4, "completions/min_terminated_length": 0.0, "entropy": 0.4293009482324123, "epoch": 6.462450592885375, "frac_reward_zero_std": 0.025, "grad_norm": 1.9479798078536987, "learning_rate": 2e-05, "loss": 0.0046630859375, "num_tokens": 97730354.0, "reward": 1.0037811279296875, "reward/completion_length_max": 41.6, "reward/completion_length_mean": 36.45, "reward/completion_length_min": 32.4, "reward/completion_length_p90": 38.4, "reward/completion_length_p95": 39.0, "reward/duplicate_query_rate": 0.03359375, "reward/duplicate_ranked_plain_query_rate": 0.03359375, "reward/invalid_output_rate": 0.05625, "reward/length_out_of_range_rate": 0.02265625, "reward/low_length_query_rate": 0.02265625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00234375, "reward/mean_colbert": 0.91875, "reward/mean_mrr": 0.94375, "reward/mean_ndcg_at_10": 0.8515625, "reward/mean_r_ndcg": 0.7453125, "reward/mean_r_ndcg_cutoff": 5.28125, "reward/mean_recall": 0.214453125, "reward/mean_recall_at_10": 0.0466796875, "reward/mean_recall_at_100": 0.214453125, "reward/mean_retrieval_combined_bounded": 0.07119140625, "reward/mean_retrieval_combined_raw": 0.17617100477218628, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.94375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.94375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1632.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.02265625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3504408061504364, "rewards/retrieval_reward/mean": 1.00234375, "rewards/retrieval_reward/std": 0.35078125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.646803283691406, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12278729081153869, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.629656410217285, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2775950461626053, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1635, "step_time": 14.12935928402003 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.11484375, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.8, "completions/min_terminated_length": 0.0, "entropy": 0.4559218347072601, "epoch": 6.482213438735178, "frac_reward_zero_std": 0.0, "grad_norm": 2.2022032737731934, "learning_rate": 2e-05, "loss": 0.00022125244140625, "num_tokens": 98034117.0, "reward": 1.02901611328125, "reward/completion_length_max": 40.4, "reward/completion_length_mean": 36.15, "reward/completion_length_min": 32.8, "reward/completion_length_p90": 38.2, "reward/completion_length_p95": 38.4, "reward/duplicate_query_rate": 0.0203125, "reward/duplicate_ranked_plain_query_rate": 0.0203125, "reward/invalid_output_rate": 0.021875, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.02890625, "reward/mean_colbert": 0.9578125, "reward/mean_mrr": 0.978125, "reward/mean_ndcg_at_10": 0.89375, "reward/mean_r_ndcg": 0.78359375, "reward/mean_r_ndcg_cutoff": 6.425, "reward/mean_recall": 0.222265625, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.222265625, "reward/mean_retrieval_combined_bounded": 0.051708984375, "reward/mean_retrieval_combined_raw": 0.08134765625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.978125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.978125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1636.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0015625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3040005207061768, "rewards/retrieval_reward/mean": 1.02890625, "rewards/retrieval_reward/std": 0.304296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.574562168121338, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.11417391896247864, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.547718906402588, "sampling/raw_vllm_sampling_logp_difference/mean": 0.297407454252243, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1640, "step_time": 14.103080415609293 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 59.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.4046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.6, "completions/min_terminated_length": 0.0, "entropy": 0.4805539928376675, "epoch": 6.5019762845849804, "frac_reward_zero_std": 0.0, "grad_norm": 2.4510748386383057, "learning_rate": 2e-05, "loss": 0.02138671875, "num_tokens": 98337995.0, "reward": 0.9943389892578125, "reward/completion_length_max": 59.2, "reward/completion_length_mean": 36.4, "reward/completion_length_min": 32.6, "reward/completion_length_p90": 38.4, "reward/completion_length_p95": 39.4, "reward/duplicate_query_rate": 0.00859375, "reward/duplicate_ranked_plain_query_rate": 0.00859375, "reward/invalid_output_rate": 0.03984375, "reward/length_out_of_range_rate": 0.02890625, "reward/low_length_query_rate": 0.02890625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.99375, "reward/mean_colbert": 0.93671875, "reward/mean_mrr": 0.959375, "reward/mean_ndcg_at_10": 0.88203125, "reward/mean_r_ndcg": 0.7671875, "reward/mean_r_ndcg_cutoff": 5.75, "reward/mean_recall": 0.20390625, "reward/mean_recall_at_10": 0.046875, "reward/mean_recall_at_100": 0.20390625, "reward/mean_retrieval_combined_bounded": 0.05318603515625, "reward/mean_retrieval_combined_raw": 0.11298208236694336, "reward/output_tag_rate": 0.99765625, "reward/query_valid_rate": 0.96015625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.96015625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1641.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.02890625, "reward/too_short_query_rate": 0.0, "reward_std": 0.38973135948181153, "rewards/retrieval_reward/mean": 0.99375, "rewards/retrieval_reward/std": 0.389453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.0286831855773926, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.09731551855802537, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.89452667236328, "sampling/raw_vllm_sampling_logp_difference/mean": 0.29446409940719603, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1645, "step_time": 15.009124238195364 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.67109375, "completions/mean_terminated_length": 0.0, "completions/min_length": 33.4, "completions/min_terminated_length": 0.0, "entropy": 0.47739302515983584, "epoch": 6.521739130434782, "frac_reward_zero_std": 0.0, "grad_norm": 2.142289638519287, "learning_rate": 2e-05, "loss": 0.004345703125, "num_tokens": 98642726.0, "reward": 1.027001953125, "reward/completion_length_max": 40.6, "reward/completion_length_mean": 36.65, "reward/completion_length_min": 33.4, "reward/completion_length_p90": 38.6, "reward/completion_length_p95": 39.0, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.02265625, "reward/length_out_of_range_rate": 0.0171875, "reward/low_length_query_rate": 0.0171875, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.95390625, "reward/mean_mrr": 0.97734375, "reward/mean_ndcg_at_10": 0.86875, "reward/mean_r_ndcg": 0.73515625, "reward/mean_r_ndcg_cutoff": 5.78125, "reward/mean_recall": 0.253125, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.253125, "reward/mean_retrieval_combined_bounded": 0.058203125, "reward/mean_retrieval_combined_raw": 0.131787109375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97734375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97734375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1648.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0171875, "reward/too_short_query_rate": 0.0, "reward_std": 0.32901719212532043, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.3296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7133780479431153, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1155370905995369, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.65978946685791, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3041897535324097, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1650, "step_time": 14.161598388198763 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 44.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 38.0703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 33.0, "completions/min_terminated_length": 0.0, "entropy": 0.45885011553764343, "epoch": 6.525691699604743, "frac_reward_zero_std": 0.0, "num_tokens": 98704088.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6774274826084263, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7284654061622441, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9024604569420035, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9947275922671354, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.5763045060257548, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6284517670797932, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8133333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 1.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3188847253632459, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.35923973893420275, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5277108433734939, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8650602409638555, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6239283717236446, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6318106895422193, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7534258581793792, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.909330985915493, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5709457294677115, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.35323050325134325, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17328337618389203, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3109475622857217, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 3.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3449.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.4210341714381211, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4570914917951953, "retrieval_eval/nq/single_turn/recall_at_10": 0.6526046196965304, "retrieval_eval/nq/single_turn/recall_at_100": 0.8878660481298927, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 1.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 299.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6724750437182632, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7032848563832259, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8344481605351171, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9374581939799331, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 0.987274169921875, "reward/completion_length_max": 44.0, "reward/completion_length_mean": 38.0, "reward/completion_length_min": 33.0, "reward/completion_length_p90": 41.0, "reward/completion_length_p95": 41.0, "reward/duplicate_query_rate": 0.05859375, "reward/duplicate_ranked_plain_query_rate": 0.05859375, "reward/invalid_output_rate": 0.14453125, "reward/length_out_of_range_rate": 0.0859375, "reward/low_length_query_rate": 0.0859375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.98828125, "reward/mean_colbert": 0.83203125, "reward/mean_mrr": 0.85546875, "reward/mean_ndcg_at_10": 0.7421875, "reward/mean_r_ndcg": 0.671875, "reward/mean_r_ndcg_cutoff": 5.625, "reward/mean_recall": 0.1884765625, "reward/mean_recall_at_10": 0.039306640625, "reward/mean_recall_at_100": 0.1884765625, "reward/mean_retrieval_combined_bounded": 0.1748046875, "reward/mean_retrieval_combined_raw": 0.400390625, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.85546875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.85546875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1648.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0859375, "reward/too_short_query_rate": 0.0, "reward_std": 0.5626057982444763, "rewards/retrieval_reward/mean": 0.98828125, "rewards/retrieval_reward/std": 0.5625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6580007076263428, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.08850957453250885, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 20.9462890625, "sampling/raw_vllm_sampling_logp_difference/mean": 0.32547953724861145, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1651, "step_time": 14.097552006947808 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 36.392578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 33.25, "completions/min_terminated_length": 0.0, "entropy": 0.493989129550755, "epoch": 6.541501976284585, "frac_reward_zero_std": 0.0, "grad_norm": 2.4136688709259033, "learning_rate": 2e-05, "loss": 0.00478515625, "num_tokens": 98947274.0, "reward": 1.0322456359863281, "reward/completion_length_max": 40.25, "reward/completion_length_mean": 36.375, "reward/completion_length_min": 33.25, "reward/completion_length_p90": 38.25, "reward/completion_length_p95": 38.75, "reward/duplicate_query_rate": 0.009765625, "reward/duplicate_ranked_plain_query_rate": 0.009765625, "reward/invalid_output_rate": 0.0146484375, "reward/length_out_of_range_rate": 0.0048828125, "reward/low_length_query_rate": 0.0048828125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.9619140625, "reward/mean_mrr": 0.9853515625, "reward/mean_ndcg_at_10": 0.88671875, "reward/mean_r_ndcg": 0.837890625, "reward/mean_r_ndcg_cutoff": 5.8125, "reward/mean_recall": 0.23291015625, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.23291015625, "reward/mean_retrieval_combined_bounded": 0.049407958984375, "reward/mean_retrieval_combined_raw": 0.09375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9853515625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9853515625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1652.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0048828125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2874305322766304, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.28759765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.2820867598056793, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.08359261602163315, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.391308546066284, "sampling/raw_vllm_sampling_logp_difference/mean": 0.31605837494134903, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1655, "step_time": 14.133448216234683 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 57.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 35.56328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 32.2, "completions/min_terminated_length": 0.0, "entropy": 0.4589768752455711, "epoch": 6.561264822134388, "frac_reward_zero_std": 0.0, "grad_norm": 2.308565378189087, "learning_rate": 2e-05, "loss": 0.006982421875, "num_tokens": 99250331.0, "reward": 1.0162933349609375, "reward/completion_length_max": 57.6, "reward/completion_length_mean": 35.55, "reward/completion_length_min": 32.2, "reward/completion_length_p90": 37.4, "reward/completion_length_p95": 38.2, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.91875, "reward/mean_r_ndcg": 0.81171875, "reward/mean_r_ndcg_cutoff": 6.60625, "reward/mean_recall": 0.26171875, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.26171875, "reward/mean_retrieval_combined_bounded": 0.0237518310546875, "reward/mean_retrieval_combined_raw": 0.04812183380126953, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1656.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.28112347722053527, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.28125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.445055389404297, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.12162977010011673, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.656537723541259, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2842021405696869, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1660, "step_time": 14.96747278557159 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.41171875, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.8, "completions/min_terminated_length": 0.0, "entropy": 0.487771362811327, "epoch": 6.5810276679841895, "frac_reward_zero_std": 0.0, "grad_norm": 1.935726284980774, "learning_rate": 2e-05, "loss": 0.002587890625, "num_tokens": 99551498.0, "reward": 1.0187255859375, "reward/completion_length_max": 40.4, "reward/completion_length_mean": 34.4, "reward/completion_length_min": 31.8, "reward/completion_length_p90": 36.4, "reward/completion_length_p95": 36.6, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.01171875, "reward/length_out_of_range_rate": 0.00390625, "reward/low_length_query_rate": 0.00390625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.9640625, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.8921875, "reward/mean_r_ndcg": 0.74375, "reward/mean_r_ndcg_cutoff": 5.46875, "reward/mean_recall": 0.220703125, "reward/mean_recall_at_10": 0.048779296875, "reward/mean_recall_at_100": 0.220703125, "reward/mean_retrieval_combined_bounded": 0.03251953125, "reward/mean_retrieval_combined_raw": 0.04115977883338928, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1664.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00390625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3026829957962036, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.303515625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.673741865158081, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.13394886553287505, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.67495412826538, "sampling/raw_vllm_sampling_logp_difference/mean": 0.28210967779159546, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1665, "step_time": 14.029348068055697 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.33359375, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.8, "completions/min_terminated_length": 0.0, "entropy": 0.4590058200061321, "epoch": 6.600790513833992, "frac_reward_zero_std": 0.0, "grad_norm": 2.3581595420837402, "learning_rate": 2e-05, "loss": -0.0007171630859375, "num_tokens": 99851669.0, "reward": 1.0207794189453125, "reward/completion_length_max": 38.2, "reward/completion_length_mean": 33.35, "reward/completion_length_min": 30.8, "reward/completion_length_p90": 35.4, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.91796875, "reward/mean_r_ndcg": 0.78671875, "reward/mean_r_ndcg_cutoff": 6.65, "reward/mean_recall": 0.2302734375, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.2302734375, "reward/mean_retrieval_combined_bounded": 0.02733154296875, "reward/mean_retrieval_combined_raw": 0.046924085915088655, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1665.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2716554760932922, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.271875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.585662603378296, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.14365388154983522, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.985724067687988, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2714109718799591, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1670, "step_time": 14.03633620843757 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 45.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.54921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.6, "completions/min_terminated_length": 0.0, "entropy": 0.39980083182454107, "epoch": 6.620553359683795, "frac_reward_zero_std": 0.0, "grad_norm": 2.840313196182251, "learning_rate": 2e-05, "loss": 0.0056396484375, "num_tokens": 100152340.0, "reward": 1.0153594970703126, "reward/completion_length_max": 45.4, "reward/completion_length_mean": 33.6, "reward/completion_length_min": 30.6, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.90234375, "reward/mean_r_ndcg": 0.8140625, "reward/mean_r_ndcg_cutoff": 6.56875, "reward/mean_recall": 0.2427734375, "reward/mean_recall_at_10": 0.0484375, "reward/mean_recall_at_100": 0.2427734375, "reward/mean_retrieval_combined_bounded": 0.02049560546875, "reward/mean_retrieval_combined_raw": 0.019595734775066376, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1672.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2855744183063507, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.2859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5567259788513184, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1457988813519478, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.808696937561034, "sampling/raw_vllm_sampling_logp_difference/mean": 0.26555308401584626, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1675, "step_time": 14.347196159348822 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.35078125, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.4, "completions/min_terminated_length": 0.0, "entropy": 0.41642596796154974, "epoch": 6.640316205533597, "frac_reward_zero_std": 0.0, "grad_norm": 2.2512664794921875, "learning_rate": 2e-05, "loss": -8.7738037109375e-05, "num_tokens": 100452309.0, "reward": 1.0174591064453125, "reward/completion_length_max": 37.4, "reward/completion_length_mean": 33.3, "reward/completion_length_min": 30.4, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.91953125, "reward/mean_r_ndcg": 0.75, "reward/mean_r_ndcg_cutoff": 5.73125, "reward/mean_recall": 0.2189453125, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.2189453125, "reward/mean_retrieval_combined_bounded": 0.0221405029296875, "reward/mean_retrieval_combined_raw": 0.030979421734809876, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1678.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.26592434346675875, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.2662109375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7852166175842283, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.19598687589168548, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.573220443725585, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24143092036247255, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1680, "step_time": 13.940400128776673 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.73515625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.0, "completions/min_terminated_length": 0.0, "entropy": 0.4155989281833172, "epoch": 6.660079051383399, "frac_reward_zero_std": 0.0, "grad_norm": 3.154083013534546, "learning_rate": 2e-05, "loss": 0.000994873046875, "num_tokens": 100751650.0, "reward": 1.0302825927734376, "reward/completion_length_max": 37.0, "reward/completion_length_mean": 32.7, "reward/completion_length_min": 30.0, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0015625, "reward/malformed_ranked_plain_query_rate": 0.0015625, "reward/mean": 1.03125, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.9078125, "reward/mean_r_ndcg": 0.8046875, "reward/mean_r_ndcg_cutoff": 5.6375, "reward/mean_recall": 0.2291015625, "reward/mean_recall_at_10": 0.049462890625, "reward/mean_recall_at_100": 0.2291015625, "reward/mean_retrieval_combined_bounded": 0.03572998046875, "reward/mean_retrieval_combined_raw": 0.1663018584251404, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1680.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.29491870403289794, "rewards/retrieval_reward/mean": 1.03125, "rewards/retrieval_reward/std": 0.295703125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6025578498840334, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15326886624097824, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.36910800933838, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2659955620765686, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1685, "step_time": 13.896515436342451 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.65, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.3618635557591915, "epoch": 6.679841897233201, "frac_reward_zero_std": 0.0, "grad_norm": 2.281200408935547, "learning_rate": 2e-05, "loss": -0.001031494140625, "num_tokens": 101049666.0, "reward": 1.0052947998046875, "reward/completion_length_max": 36.2, "reward/completion_length_mean": 31.65, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 33.8, "reward/completion_length_p95": 34.6, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00390625, "reward/mean_colbert": 0.97578125, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.91953125, "reward/mean_r_ndcg": 0.75, "reward/mean_r_ndcg_cutoff": 6.35, "reward/mean_recall": 0.2091796875, "reward/mean_recall_at_10": 0.049658203125, "reward/mean_recall_at_100": 0.2091796875, "reward/mean_retrieval_combined_bounded": 0.0060546875, "reward/mean_retrieval_combined_raw": -0.03090902790427208, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1688.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2733790874481201, "rewards/retrieval_reward/mean": 1.00390625, "rewards/retrieval_reward/std": 0.2734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.726812219619751, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.26379466652870176, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.915436744689941, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20101031064987182, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1690, "step_time": 13.819699119403959 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.90546875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.2, "completions/min_terminated_length": 0.0, "entropy": 0.3741402685642242, "epoch": 6.699604743083004, "frac_reward_zero_std": 0.0, "grad_norm": 3.0654819011688232, "learning_rate": 2e-05, "loss": 3.4332275390625e-05, "num_tokens": 101346889.0, "reward": 1.0132965087890624, "reward/completion_length_max": 34.6, "reward/completion_length_mean": 30.9, "reward/completion_length_min": 28.2, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 33.4, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.8890625, "reward/mean_r_ndcg": 0.7796875, "reward/mean_r_ndcg_cutoff": 7.35, "reward/mean_recall": 0.218359375, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.218359375, "reward/mean_retrieval_combined_bounded": 0.01949462890625, "reward/mean_retrieval_combined_raw": 0.026365846395492554, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1692.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2967900216579437, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.298046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6175310134887697, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.22108905017375946, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.132074356079102, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24218326508998872, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1695, "step_time": 13.757707593997475 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.696875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.4, "completions/min_terminated_length": 0.0, "entropy": 0.3849832832813263, "epoch": 6.719367588932807, "frac_reward_zero_std": 0.0, "grad_norm": 2.684859037399292, "learning_rate": 2e-05, "loss": 0.000982666015625, "num_tokens": 101643589.0, "reward": 1.02071533203125, "reward/completion_length_max": 35.0, "reward/completion_length_mean": 30.7, "reward/completion_length_min": 27.4, "reward/completion_length_p90": 32.8, "reward/completion_length_p95": 33.2, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.9046875, "reward/mean_r_ndcg": 0.75859375, "reward/mean_r_ndcg_cutoff": 5.63125, "reward/mean_recall": 0.225, "reward/mean_recall_at_10": 0.048974609375, "reward/mean_recall_at_100": 0.225, "reward/mean_retrieval_combined_bounded": 0.0278076171875, "reward/mean_retrieval_combined_raw": 0.04023517966270447, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1696.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.26213141083717345, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.2623046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7884684085845945, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.24171265661716462, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.284170150756836, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22290563881397246, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1700, "step_time": 13.808878905547317 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.9921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.33170562610030174, "epoch": 6.723320158102767, "frac_reward_zero_std": 0.0, "num_tokens": 101703171.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6713044453440432, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7237551899557688, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9024604569420035, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9938488576449912, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6204675950389625, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.670739155257204, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8466666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3236010063492542, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.36047753660322895, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5204819277108433, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8530120481927711, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6157403500220378, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6245257716849298, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7487310224986281, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9160588531187123, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5780120213434803, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3607807839293317, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17602292633549144, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.31553588992095993, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 2.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3450.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.414704198755656, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4506345821807987, "retrieval_eval/nq/single_turn/recall_at_10": 0.6480676328502416, "retrieval_eval/nq/single_turn/recall_at_100": 0.8887198067632851, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6509075029009911, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6818411454495357, "retrieval_eval/scifact/single_turn/recall_at_10": 0.82, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9560000000000001, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 0.992706298828125, "reward/completion_length_max": 37.0, "reward/completion_length_mean": 31.0, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.9921875, "reward/mean_colbert": 0.9765625, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.8125, "reward/mean_r_ndcg": 0.68359375, "reward/mean_r_ndcg_cutoff": 6.5, "reward/mean_recall": 0.27734375, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.27734375, "reward/mean_retrieval_combined_bounded": -0.00738525390625, "reward/mean_retrieval_combined_raw": 2.2817403078079224e-07, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1696.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.31185418367385864, "rewards/retrieval_reward/mean": 0.9921875, "rewards/retrieval_reward/std": 0.3125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.573324203491211, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.30144840478897095, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 19.174610137939453, "sampling/raw_vllm_sampling_logp_difference/mean": 0.3302411437034607, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1701, "step_time": 13.942961885943078 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.9794921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.3503044843673706, "epoch": 6.739130434782608, "frac_reward_zero_std": 0.0, "grad_norm": 2.0588173866271973, "learning_rate": 2e-05, "loss": 3.337860107421875e-05, "num_tokens": 101941038.0, "reward": 1.0014762878417969, "reward/completion_length_max": 35.0, "reward/completion_length_mean": 30.96875, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 33.75, "reward/duplicate_query_rate": 0.0234375, "reward/duplicate_ranked_plain_query_rate": 0.0234375, "reward/invalid_output_rate": 0.0234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0029296875, "reward/mean_colbert": 0.953125, "reward/mean_mrr": 0.9765625, "reward/mean_ndcg_at_10": 0.8671875, "reward/mean_r_ndcg": 0.7177734375, "reward/mean_r_ndcg_cutoff": 6.28125, "reward/mean_recall": 0.258056640625, "reward/mean_recall_at_10": 0.04876708984375, "reward/mean_recall_at_100": 0.258056640625, "reward/mean_retrieval_combined_bounded": 0.02501678466796875, "reward/mean_retrieval_combined_raw": 0.039663225412368774, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1704.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2922201231122017, "rewards/retrieval_reward/mean": 1.0029296875, "rewards/retrieval_reward/std": 0.29248046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.792982041835785, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2614111565053463, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.425751447677612, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20341335982084274, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1705, "step_time": 13.718277961219428 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.484375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.3238479383289814, "epoch": 6.758893280632411, "frac_reward_zero_std": 0.0, "grad_norm": 2.147597074508667, "learning_rate": 2e-05, "loss": -0.002001953125, "num_tokens": 102237402.0, "reward": 1.00797119140625, "reward/completion_length_max": 33.8, "reward/completion_length_mean": 30.475, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 32.2, "reward/completion_length_p95": 33.2, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.009375, "reward/mean_colbert": 0.97578125, "reward/mean_mrr": 0.99765625, "reward/mean_ndcg_at_10": 0.91640625, "reward/mean_r_ndcg": 0.8328125, "reward/mean_r_ndcg_cutoff": 6.6875, "reward/mean_recall": 0.241015625, "reward/mean_recall_at_10": 0.04892578125, "reward/mean_recall_at_100": 0.241015625, "reward/mean_retrieval_combined_bounded": 0.010345458984375, "reward/mean_retrieval_combined_raw": -0.06142539456486702, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1707.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2939544439315796, "rewards/retrieval_reward/mean": 1.009375, "rewards/retrieval_reward/std": 0.29453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8000502586364746, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.28567902445793153, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.223550033569335, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20278393030166625, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1710, "step_time": 13.731475630030037 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.30234375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.4098734758794308, "epoch": 6.778656126482213, "frac_reward_zero_std": 0.0, "grad_norm": 2.0089221000671387, "learning_rate": 2e-05, "loss": 0.000335693359375, "num_tokens": 102533629.0, "reward": 1.025433349609375, "reward/completion_length_max": 34.4, "reward/completion_length_mean": 30.3, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 32.2, "reward/completion_length_p95": 32.6, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.975, "reward/mean_mrr": 0.98125, "reward/mean_ndcg_at_10": 0.8953125, "reward/mean_r_ndcg": 0.76640625, "reward/mean_r_ndcg_cutoff": 5.89375, "reward/mean_recall": 0.215234375, "reward/mean_recall_at_10": 0.047607421875, "reward/mean_recall_at_100": 0.215234375, "reward/mean_retrieval_combined_bounded": 0.026363372802734375, "reward/mean_retrieval_combined_raw": 0.08037300258874894, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1712.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2797569394111633, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.2796875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.777093458175659, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2519564390182495, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.945566463470459, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20651786625385285, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1715, "step_time": 13.789224326203112 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.8375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.39549518972635267, "epoch": 6.798418972332016, "frac_reward_zero_std": 0.0, "grad_norm": 2.7128758430480957, "learning_rate": 2e-05, "loss": 0.000653076171875, "num_tokens": 102830413.0, "reward": 1.0196868896484375, "reward/completion_length_max": 34.4, "reward/completion_length_mean": 30.825, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 33.2, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.9140625, "reward/mean_r_ndcg": 0.8171875, "reward/mean_r_ndcg_cutoff": 6.89375, "reward/mean_recall": 0.2376953125, "reward/mean_recall_at_10": 0.048876953125, "reward/mean_recall_at_100": 0.2376953125, "reward/mean_retrieval_combined_bounded": 0.0236083984375, "reward/mean_retrieval_combined_raw": 0.023318296670913695, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1716.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27870320081710814, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.27890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.671688508987427, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.24732151627540588, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.7672216415405275, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21109262704849244, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1720, "step_time": 13.731237244384829 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.8875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.8, "completions/min_terminated_length": 0.0, "entropy": 0.3910932280123234, "epoch": 6.818181818181818, "frac_reward_zero_std": 0.0, "grad_norm": 2.473217725753784, "learning_rate": 2e-05, "loss": 0.00024871826171875, "num_tokens": 103127421.0, "reward": 1.0147430419921875, "reward/completion_length_max": 35.4, "reward/completion_length_mean": 30.875, "reward/completion_length_min": 27.8, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 33.4, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01328125, "reward/mean_colbert": 0.9765625, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.90703125, "reward/mean_r_ndcg": 0.81796875, "reward/mean_r_ndcg_cutoff": 6.15, "reward/mean_recall": 0.2533203125, "reward/mean_recall_at_10": 0.049365234375, "reward/mean_recall_at_100": 0.2533203125, "reward/mean_retrieval_combined_bounded": 0.014697265625, "reward/mean_retrieval_combined_raw": 7.429439574480057e-06, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1721.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2710203528404236, "rewards/retrieval_reward/mean": 1.01328125, "rewards/retrieval_reward/std": 0.271484375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7738900184631348, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20795702636241914, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.782212448120116, "sampling/raw_vllm_sampling_logp_difference/mean": 0.25175015032291415, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1725, "step_time": 13.822351316793355 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.48203125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.39347644001245496, "epoch": 6.837944664031621, "frac_reward_zero_std": 0.0, "grad_norm": 2.330024003982544, "learning_rate": 2e-05, "loss": -0.00208740234375, "num_tokens": 103425222.0, "reward": 1.014678955078125, "reward/completion_length_max": 36.6, "reward/completion_length_mean": 31.45, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 33.8, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.975, "reward/mean_mrr": 0.99765625, "reward/mean_ndcg_at_10": 0.92734375, "reward/mean_r_ndcg": 0.8015625, "reward/mean_r_ndcg_cutoff": 6.3875, "reward/mean_recall": 0.230078125, "reward/mean_recall_at_10": 0.04921875, "reward/mean_recall_at_100": 0.230078125, "reward/mean_retrieval_combined_bounded": 0.017138671875, "reward/mean_retrieval_combined_raw": 0.018991343677043915, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1728.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.26698005199432373, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.266796875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7147384166717528, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.20280361473560332, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.253266048431396, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24385570287704467, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1730, "step_time": 13.854805516579654 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.60625, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.3917235754430294, "epoch": 6.857707509881423, "frac_reward_zero_std": 0.0, "grad_norm": 2.310513973236084, "learning_rate": 2e-05, "loss": 0.000531005859375, "num_tokens": 103722958.0, "reward": 1.0254119873046874, "reward/completion_length_max": 36.2, "reward/completion_length_mean": 31.6, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 33.6, "reward/completion_length_p95": 34.2, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.8890625, "reward/mean_r_ndcg": 0.7203125, "reward/mean_r_ndcg_cutoff": 6.63125, "reward/mean_recall": 0.2630859375, "reward/mean_recall_at_10": 0.048193359375, "reward/mean_recall_at_100": 0.2630859375, "reward/mean_retrieval_combined_bounded": 0.0327880859375, "reward/mean_retrieval_combined_raw": 0.03178741782903671, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1731.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2883664906024933, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.288671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.694879674911499, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1908068150281906, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.89419116973877, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2436820536851883, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1735, "step_time": 13.859012293827254 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.82265625, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.38703876957297323, "epoch": 6.877470355731226, "frac_reward_zero_std": 0.0, "grad_norm": 2.202171564102173, "learning_rate": 2e-05, "loss": -3.8909912109375e-05, "num_tokens": 104021131.0, "reward": 1.018804931640625, "reward/completion_length_max": 35.2, "reward/completion_length_mean": 31.825, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01796875, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.90234375, "reward/mean_r_ndcg": 0.81171875, "reward/mean_r_ndcg_cutoff": 6.4, "reward/mean_recall": 0.22890625, "reward/mean_recall_at_10": 0.049365234375, "reward/mean_recall_at_100": 0.22890625, "reward/mean_retrieval_combined_bounded": 0.0234619140625, "reward/mean_retrieval_combined_raw": 0.02548692226409912, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1736.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3022913336753845, "rewards/retrieval_reward/mean": 1.01796875, "rewards/retrieval_reward/std": 0.302734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.81473593711853, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21983564496040345, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.317586708068848, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21812937557697296, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1740, "step_time": 13.841349015396554 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.0828125, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.2, "completions/min_terminated_length": 0.0, "entropy": 0.4000122509896755, "epoch": 6.897233201581027, "frac_reward_zero_std": 0.0, "grad_norm": 1.9113572835922241, "learning_rate": 2e-05, "loss": 8.65936279296875e-05, "num_tokens": 104321045.0, "reward": 1.0259796142578126, "reward/completion_length_max": 37.6, "reward/completion_length_mean": 33.05, "reward/completion_length_min": 30.2, "reward/completion_length_p90": 34.8, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0265625, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.928125, "reward/mean_r_ndcg": 0.82890625, "reward/mean_r_ndcg_cutoff": 6.85, "reward/mean_recall": 0.2337890625, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.2337890625, "reward/mean_retrieval_combined_bounded": 0.02999267578125, "reward/mean_retrieval_combined_raw": 0.03178618107922375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1744.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2621509194374084, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.2623046875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.870733690261841, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.24764698147773742, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.82283296585083, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22004097402095796, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1745, "step_time": 13.971288509806618 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.62734375, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.6, "completions/min_terminated_length": 0.0, "entropy": 0.4441208191215992, "epoch": 6.91699604743083, "frac_reward_zero_std": 0.0, "grad_norm": 2.091336488723755, "learning_rate": 2e-05, "loss": -0.001129150390625, "num_tokens": 104621432.0, "reward": 1.01634521484375, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 33.55, "reward/completion_length_min": 30.6, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.88828125, "reward/mean_r_ndcg": 0.78828125, "reward/mean_r_ndcg_cutoff": 5.1, "reward/mean_recall": 0.2203125, "reward/mean_recall_at_10": 0.048046875, "reward/mean_recall_at_100": 0.2203125, "reward/mean_retrieval_combined_bounded": 0.0226470947265625, "reward/mean_retrieval_combined_raw": 0.03210373353213072, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1745.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2730883061885834, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.2734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.786763858795166, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2287381559610367, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.631053638458251, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22615952491760255, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1750, "step_time": 13.938141247816384 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.875, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.0, "completions/min_terminated_length": 0.0, "entropy": 0.435627106577158, "epoch": 6.920948616600791, "frac_reward_zero_std": 0.0, "num_tokens": 104681464.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6625104089039324, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7175617170966706, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9042179261862917, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9973637961335676, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6266527889925743, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6747570995381019, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8466666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3268247973645679, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3650618388199188, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5301204819277109, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8554216867469879, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.638329707617728, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6473448774563747, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7775204255838059, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9115107310529845, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5837209106856359, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3639198004135322, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.1802542250039079, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3122522431214195, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 0.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3452.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.4160353091933557, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.45447297063203274, "retrieval_eval/nq/single_turn/recall_at_10": 0.6568655851680185, "retrieval_eval/nq/single_turn/recall_at_100": 0.8958574739281576, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6746529481105842, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7089427968331664, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8506666666666667, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9553333333333334, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0229339599609375, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 34.0, "reward/completion_length_min": 31.0, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.8828125, "reward/mean_r_ndcg": 0.671875, "reward/mean_r_ndcg_cutoff": 4.84375, "reward/mean_recall": 0.271484375, "reward/mean_recall_at_10": 0.0498046875, "reward/mean_recall_at_100": 0.271484375, "reward/mean_retrieval_combined_bounded": 0.0269775390625, "reward/mean_retrieval_combined_raw": 0.0272216796875, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1752.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27968984842300415, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.28125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8318638801574707, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21668703854084015, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.468751907348633, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27768343687057495, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1751, "step_time": 13.93499069812242 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.75, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.69140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.75, "completions/min_terminated_length": 0.0, "entropy": 0.41804615408182144, "epoch": 6.936758893280633, "frac_reward_zero_std": 0.0, "grad_norm": 2.282802104949951, "learning_rate": 2e-05, "loss": 0.001202392578125, "num_tokens": 104921884.0, "reward": 1.0187759399414062, "reward/completion_length_max": 38.75, "reward/completion_length_mean": 33.6875, "reward/completion_length_min": 30.75, "reward/completion_length_p90": 36.0, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.015625, "reward/duplicate_ranked_plain_query_rate": 0.015625, "reward/invalid_output_rate": 0.015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01953125, "reward/mean_colbert": 0.9609375, "reward/mean_mrr": 0.984375, "reward/mean_ndcg_at_10": 0.908203125, "reward/mean_r_ndcg": 0.828125, "reward/mean_r_ndcg_cutoff": 6.1640625, "reward/mean_recall": 0.208251953125, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.208251953125, "reward/mean_retrieval_combined_bounded": 0.03424072265625, "reward/mean_retrieval_combined_raw": 0.05359084904193878, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1752.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2818589359521866, "rewards/retrieval_reward/mean": 1.01953125, "rewards/retrieval_reward/std": 0.28271484375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7368812561035156, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.19473832473158836, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.354153633117676, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21682432666420937, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1755, "step_time": 13.95614599547116 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.2140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.2, "completions/min_terminated_length": 0.0, "entropy": 0.4025729849934578, "epoch": 6.956521739130435, "frac_reward_zero_std": 0.0, "grad_norm": 2.2568178176879883, "learning_rate": 2e-05, "loss": -0.000823974609375, "num_tokens": 105221934.0, "reward": 1.0207489013671875, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 33.2, "reward/completion_length_min": 30.2, "reward/completion_length_p90": 35.0, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0203125, "reward/mean_colbert": 0.975, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.92421875, "reward/mean_r_ndcg": 0.83359375, "reward/mean_r_ndcg_cutoff": 7.0125, "reward/mean_recall": 0.2548828125, "reward/mean_recall_at_10": 0.049462890625, "reward/mean_recall_at_100": 0.2548828125, "reward/mean_retrieval_combined_bounded": 0.02236328125, "reward/mean_retrieval_combined_raw": 0.012479256838560104, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1758.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2909191608428955, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.291796875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.589944267272949, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.19547105729579925, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.555096244812011, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24066925048828125, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1760, "step_time": 13.906514453806448 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.56796875, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.8, "completions/min_terminated_length": 0.0, "entropy": 0.41492756456136703, "epoch": 6.976284584980237, "frac_reward_zero_std": 0.0, "grad_norm": 3.253304958343506, "learning_rate": 2e-05, "loss": -0.000421142578125, "num_tokens": 105522213.0, "reward": 1.007086181640625, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 33.55, "reward/completion_length_min": 30.8, "reward/completion_length_p90": 35.4, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.01875, "reward/duplicate_ranked_plain_query_rate": 0.01875, "reward/invalid_output_rate": 0.01875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00703125, "reward/mean_colbert": 0.95625, "reward/mean_mrr": 0.98125, "reward/mean_ndcg_at_10": 0.87109375, "reward/mean_r_ndcg": 0.72421875, "reward/mean_r_ndcg_cutoff": 6.06875, "reward/mean_recall": 0.2287109375, "reward/mean_recall_at_10": 0.048095703125, "reward/mean_recall_at_100": 0.2287109375, "reward/mean_retrieval_combined_bounded": 0.0260009765625, "reward/mean_retrieval_combined_raw": 0.051904449611902236, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1760.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3159246504306793, "rewards/retrieval_reward/mean": 1.00703125, "rewards/retrieval_reward/std": 0.316796875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.847370719909668, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1754411593079567, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.470685958862305, "sampling/raw_vllm_sampling_logp_difference/mean": 0.266928905248642, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1765, "step_time": 13.861460603808519 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 34.2, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.2, "completions/min_terminated_length": 0.0, "entropy": 0.41338020786643026, "epoch": 6.996047430830039, "frac_reward_zero_std": 0.0, "grad_norm": 1.7605829238891602, "learning_rate": 2e-05, "loss": 0.00012969970703125, "num_tokens": 105823493.0, "reward": 1.029022216796875, "reward/completion_length_max": 39.6, "reward/completion_length_mean": 34.2, "reward/completion_length_min": 31.2, "reward/completion_length_p90": 36.4, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0296875, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.89609375, "reward/mean_r_ndcg": 0.77109375, "reward/mean_r_ndcg_cutoff": 6.69375, "reward/mean_recall": 0.235546875, "reward/mean_recall_at_10": 0.048388671875, "reward/mean_recall_at_100": 0.235546875, "reward/mean_retrieval_combined_bounded": 0.0340576171875, "reward/mean_retrieval_combined_raw": 0.02351827141828835, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1768.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.28358414173126223, "rewards/retrieval_reward/mean": 1.028125, "rewards/retrieval_reward/std": 0.28359375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6498762130737306, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.19852295964956285, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.57692584991455, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23625852167606354, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1770, "step_time": 14.043672201992013 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.95625, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.4, "completions/min_terminated_length": 0.0, "entropy": 0.39974283427000046, "epoch": 7.015810276679842, "frac_reward_zero_std": 0.0, "grad_norm": 2.288818120956421, "learning_rate": 2e-05, "loss": -0.00072021484375, "num_tokens": 106124333.0, "reward": 0.9898834228515625, "reward/completion_length_max": 39.6, "reward/completion_length_mean": 33.9, "reward/completion_length_min": 30.4, "reward/completion_length_p90": 37.0, "reward/completion_length_p95": 37.4, "reward/duplicate_query_rate": 0.0203125, "reward/duplicate_ranked_plain_query_rate": 0.0203125, "reward/invalid_output_rate": 0.02265625, "reward/length_out_of_range_rate": 0.00234375, "reward/low_length_query_rate": 0.00234375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 0.9890625, "reward/mean_colbert": 0.953125, "reward/mean_mrr": 0.97734375, "reward/mean_ndcg_at_10": 0.9015625, "reward/mean_r_ndcg": 0.80859375, "reward/mean_r_ndcg_cutoff": 6.70625, "reward/mean_recall": 0.2349609375, "reward/mean_recall_at_10": 0.048388671875, "reward/mean_recall_at_100": 0.2349609375, "reward/mean_retrieval_combined_bounded": 0.013623046875, "reward/mean_retrieval_combined_raw": -0.027225685119628907, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.97734375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.97734375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1772.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00234375, "reward/too_short_query_rate": 0.0, "reward_std": 0.31621202230453493, "rewards/retrieval_reward/mean": 0.9890625, "rewards/retrieval_reward/std": 0.31640625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.796047830581665, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.25037859082221986, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.737033557891845, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22215254306793214, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1775, "step_time": 14.037616013386288 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 40.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.834375, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.2, "completions/min_terminated_length": 0.0, "entropy": 0.3948855519294739, "epoch": 7.0355731225296445, "frac_reward_zero_std": 0.0, "grad_norm": 1.8146902322769165, "learning_rate": 2e-05, "loss": 0.0013916015625, "num_tokens": 106424921.0, "reward": 1.024114990234375, "reward/completion_length_max": 40.4, "reward/completion_length_mean": 33.85, "reward/completion_length_min": 30.2, "reward/completion_length_p90": 36.2, "reward/completion_length_p95": 37.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.003125, "reward/low_length_query_rate": 0.003125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.896875, "reward/mean_r_ndcg": 0.7421875, "reward/mean_r_ndcg_cutoff": 6.68125, "reward/mean_recall": 0.2310546875, "reward/mean_recall_at_10": 0.048779296875, "reward/mean_recall_at_100": 0.2310546875, "reward/mean_retrieval_combined_bounded": 0.02891845703125, "reward/mean_retrieval_combined_raw": 0.05717147886753082, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1776.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.003125, "reward/too_short_query_rate": 0.0, "reward_std": 0.2924624502658844, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.29296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.793743944168091, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2712433934211731, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.577322578430175, "sampling/raw_vllm_sampling_logp_difference/mean": 0.19405255019664763, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1780, "step_time": 14.101456326351036 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 56.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.4453125, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.2, "completions/min_terminated_length": 0.0, "entropy": 0.43069804459810257, "epoch": 7.055335968379446, "frac_reward_zero_std": 0.0, "grad_norm": 2.341324806213379, "learning_rate": 2e-05, "loss": 0.01416015625, "num_tokens": 106725331.0, "reward": 1.013702392578125, "reward/completion_length_max": 56.2, "reward/completion_length_mean": 33.5, "reward/completion_length_min": 30.2, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 36.4, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.00625, "reward/low_length_query_rate": 0.00625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.9671875, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.90703125, "reward/mean_r_ndcg": 0.79296875, "reward/mean_r_ndcg_cutoff": 6.475, "reward/mean_recall": 0.228515625, "reward/mean_recall_at_10": 0.049267578125, "reward/mean_recall_at_100": 0.228515625, "reward/mean_retrieval_combined_bounded": 0.026318359375, "reward/mean_retrieval_combined_raw": 0.02882925271987915, "reward/output_tag_rate": 0.99921875, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1782.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00625, "reward/too_short_query_rate": 0.0, "reward_std": 0.3008813142776489, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.3015625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.732833242416382, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.24810679852962494, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.665735340118408, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21925226747989654, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1785, "step_time": 14.784347501769663 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.67421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 30.4, "completions/min_terminated_length": 0.0, "entropy": 0.43997653275728227, "epoch": 7.075098814229249, "frac_reward_zero_std": 0.0, "grad_norm": 1.576393723487854, "learning_rate": 2e-05, "loss": -0.001019287109375, "num_tokens": 107026290.0, "reward": 1.0078704833984375, "reward/completion_length_max": 37.8, "reward/completion_length_mean": 33.6, "reward/completion_length_min": 30.4, "reward/completion_length_p90": 35.6, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.009375, "reward/mean_colbert": 0.97421875, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.93515625, "reward/mean_r_ndcg": 0.78515625, "reward/mean_r_ndcg_cutoff": 6.84375, "reward/mean_recall": 0.220703125, "reward/mean_recall_at_10": 0.049267578125, "reward/mean_recall_at_100": 0.220703125, "reward/mean_retrieval_combined_bounded": 0.008655548095703125, "reward/mean_retrieval_combined_raw": 0.006243984028697014, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1787.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2876451313495636, "rewards/retrieval_reward/mean": 1.009375, "rewards/retrieval_reward/std": 0.287890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.402921199798584, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.17963694930076599, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.498330879211426, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27052333056926725, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1790, "step_time": 13.958664232399315 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.5046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 31.0, "completions/min_terminated_length": 0.0, "entropy": 0.41652774810791016, "epoch": 7.094861660079052, "frac_reward_zero_std": 0.0, "grad_norm": 2.076097249984741, "learning_rate": 2e-05, "loss": -0.0004638671875, "num_tokens": 107326904.0, "reward": 1.0112762451171875, "reward/completion_length_max": 37.4, "reward/completion_length_mean": 33.5, "reward/completion_length_min": 31.0, "reward/completion_length_p90": 35.8, "reward/completion_length_p95": 36.0, "reward/duplicate_query_rate": 0.015625, "reward/duplicate_ranked_plain_query_rate": 0.015625, "reward/invalid_output_rate": 0.015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01015625, "reward/mean_colbert": 0.95859375, "reward/mean_mrr": 0.984375, "reward/mean_ndcg_at_10": 0.88203125, "reward/mean_r_ndcg": 0.73828125, "reward/mean_r_ndcg_cutoff": 5.63125, "reward/mean_recall": 0.228125, "reward/mean_recall_at_10": 0.04775390625, "reward/mean_recall_at_100": 0.228125, "reward/mean_retrieval_combined_bounded": 0.02698974609375, "reward/mean_retrieval_combined_raw": 0.05112612247467041, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1792.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2985089421272278, "rewards/retrieval_reward/mean": 1.01015625, "rewards/retrieval_reward/std": 0.2990234375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.68675537109375, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.18272005319595336, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 14.681807518005371, "sampling/raw_vllm_sampling_logp_difference/mean": 0.254985636472702, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1795, "step_time": 13.894092906254809 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 37.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 33.07578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.8, "completions/min_terminated_length": 0.0, "entropy": 0.42107178941369056, "epoch": 7.1146245059288535, "frac_reward_zero_std": 0.0, "grad_norm": 2.8956480026245117, "learning_rate": 2e-05, "loss": 0.00013427734375, "num_tokens": 107626265.0, "reward": 1.0188507080078124, "reward/completion_length_max": 37.2, "reward/completion_length_mean": 33.1, "reward/completion_length_min": 29.8, "reward/completion_length_p90": 35.2, "reward/completion_length_p95": 35.8, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.97578125, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.93125, "reward/mean_r_ndcg": 0.83125, "reward/mean_r_ndcg_cutoff": 6.15, "reward/mean_recall": 0.2181640625, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.2181640625, "reward/mean_retrieval_combined_bounded": 0.0189208984375, "reward/mean_retrieval_combined_raw": -1.675933599472046e-05, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1796.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.26909510493278505, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.269140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.5309309005737304, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1990973800420761, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.071062088012695, "sampling/raw_vllm_sampling_logp_difference/mean": 0.24000315368175507, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1800, "step_time": 13.884638131188694 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 38.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.69921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.4077664203941822, "epoch": 7.118577075098814, "frac_reward_zero_std": 0.0, "num_tokens": 107685996.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6669737909795983, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7213673778592786, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9050966608084359, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9956063268892794, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.604181974701912, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6625502132472004, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.86, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3366670101218745, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.37271826697847654, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5301204819277109, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8506024096385543, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6323594000556719, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.641404335534573, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.763549097615999, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9067530181086519, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 1.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 322.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5837422095050113, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.36822777012476215, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.18178060295387213, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3113236467304932, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 1.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3451.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.4224313851700962, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.45774281299363084, "retrieval_eval/nq/single_turn/recall_at_10": 0.6527093596059114, "retrieval_eval/nq/single_turn/recall_at_100": 0.8937988988698928, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6567903347169811, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6885852890069116, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8266666666666667, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9486666666666668, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.025360107421875, "reward/completion_length_max": 38.0, "reward/completion_length_mean": 32.75, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.9765625, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.9453125, "reward/mean_r_ndcg": 0.9375, "reward/mean_r_ndcg_cutoff": 5.875, "reward/mean_recall": 0.2236328125, "reward/mean_recall_at_10": 0.047607421875, "reward/mean_recall_at_100": 0.2236328125, "reward/mean_retrieval_combined_bounded": 0.0252685546875, "reward/mean_retrieval_combined_raw": 2.372264862060547e-05, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1800.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2385411560535431, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.2392578125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7917749881744385, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.15360240638256073, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.984519958496094, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2580806612968445, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1801, "step_time": 13.81626391102327 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.44921875, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.75, "completions/min_terminated_length": 0.0, "entropy": 0.3597492938861251, "epoch": 7.134387351778656, "frac_reward_zero_std": 0.0, "grad_norm": 2.4332385063171387, "learning_rate": 2e-05, "loss": -0.001422119140625, "num_tokens": 107924984.0, "reward": 1.0132217407226562, "reward/completion_length_max": 36.0, "reward/completion_length_mean": 32.4375, "reward/completion_length_min": 29.75, "reward/completion_length_p90": 34.0, "reward/completion_length_p95": 34.5, "reward/duplicate_query_rate": 0.0009765625, "reward/duplicate_ranked_plain_query_rate": 0.0009765625, "reward/invalid_output_rate": 0.0009765625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.013671875, "reward/mean_colbert": 0.9755859375, "reward/mean_mrr": 0.9970703125, "reward/mean_ndcg_at_10": 0.91015625, "reward/mean_r_ndcg": 0.8046875, "reward/mean_r_ndcg_cutoff": 5.9375, "reward/mean_recall": 0.20849609375, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.20849609375, "reward/mean_retrieval_combined_bounded": 0.014239788055419922, "reward/mean_retrieval_combined_raw": 0.008000537753105164, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9990234375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9990234375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1802.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28848933428525925, "rewards/retrieval_reward/mean": 1.013671875, "rewards/retrieval_reward/std": 0.2890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.74289733171463, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21051174402236938, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 12.739361047744751, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2797514423727989, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1805, "step_time": 13.823149817006197 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.8375, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.6, "completions/min_terminated_length": 0.0, "entropy": 0.40141771361231804, "epoch": 7.154150197628459, "frac_reward_zero_std": 0.0, "grad_norm": 2.1177139282226562, "learning_rate": 2e-05, "loss": -0.001104736328125, "num_tokens": 108223304.0, "reward": 1.011761474609375, "reward/completion_length_max": 34.8, "reward/completion_length_mean": 31.85, "reward/completion_length_min": 29.6, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.6, "reward/duplicate_query_rate": 0.00625, "reward/duplicate_ranked_plain_query_rate": 0.00625, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01171875, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.9109375, "reward/mean_r_ndcg": 0.7953125, "reward/mean_r_ndcg_cutoff": 6.275, "reward/mean_recall": 0.21953125, "reward/mean_recall_at_10": 0.04814453125, "reward/mean_recall_at_100": 0.21953125, "reward/mean_retrieval_combined_bounded": 0.01807861328125, "reward/mean_retrieval_combined_raw": -0.036119604110717775, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1808.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28498442769050597, "rewards/retrieval_reward/mean": 1.01171875, "rewards/retrieval_reward/std": 0.285546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8597299575805666, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.21440306305885315, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.680041694641114, "sampling/raw_vllm_sampling_logp_difference/mean": 0.27464519143104554, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1810, "step_time": 13.802443841006607 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.68125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.2, "completions/min_terminated_length": 0.0, "entropy": 0.3734986767172813, "epoch": 7.173913043478261, "frac_reward_zero_std": 0.0, "grad_norm": 2.627365827560425, "learning_rate": 2e-05, "loss": 0.0003509521484375, "num_tokens": 108521136.0, "reward": 1.018121337890625, "reward/completion_length_max": 35.6, "reward/completion_length_mean": 31.675, "reward/completion_length_min": 29.2, "reward/completion_length_p90": 33.4, "reward/completion_length_p95": 33.4, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.8828125, "reward/mean_r_ndcg": 0.74765625, "reward/mean_r_ndcg_cutoff": 4.815625, "reward/mean_recall": 0.2130859375, "reward/mean_recall_at_10": 0.0484375, "reward/mean_recall_at_100": 0.2130859375, "reward/mean_retrieval_combined_bounded": 0.01953582763671875, "reward/mean_retrieval_combined_raw": 0.012346112728118896, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1811.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28105091750621797, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.280859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.84716854095459, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.1968804568052292, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.72802505493164, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2594791650772095, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1815, "step_time": 14.322124686616007 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.30078125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.3769501931965351, "epoch": 7.1936758893280635, "frac_reward_zero_std": 0.0, "grad_norm": 2.345067262649536, "learning_rate": 2e-05, "loss": 0.00260009765625, "num_tokens": 108818833.0, "reward": 1.022845458984375, "reward/completion_length_max": 35.2, "reward/completion_length_mean": 31.275, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 33.4, "reward/completion_length_p95": 33.8, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.88828125, "reward/mean_r_ndcg": 0.72265625, "reward/mean_r_ndcg_cutoff": 5.925, "reward/mean_recall": 0.2138671875, "reward/mean_recall_at_10": 0.048388671875, "reward/mean_recall_at_100": 0.2138671875, "reward/mean_retrieval_combined_bounded": 0.02989501953125, "reward/mean_retrieval_combined_raw": 0.04248022437095642, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1816.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.26346365213394163, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.263671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.815873861312866, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.23443998098373414, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 11.235693168640136, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2171210914850235, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1820, "step_time": 14.530294572596905 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.97578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.3511250264942646, "epoch": 7.213438735177865, "frac_reward_zero_std": 0.0, "grad_norm": 2.9248287677764893, "learning_rate": 2e-05, "loss": 3.16619873046875e-05, "num_tokens": 109115794.0, "reward": 1.024896240234375, "reward/completion_length_max": 34.2, "reward/completion_length_mean": 30.975, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.00625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.00078125, "reward/malformed_ranked_plain_query_rate": 0.00078125, "reward/mean": 1.0265625, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.99375, "reward/mean_ndcg_at_10": 0.89453125, "reward/mean_r_ndcg": 0.809375, "reward/mean_r_ndcg_cutoff": 6.01875, "reward/mean_recall": 0.2103515625, "reward/mean_recall_at_10": 0.048291015625, "reward/mean_recall_at_100": 0.2103515625, "reward/mean_retrieval_combined_bounded": 0.03189697265625, "reward/mean_retrieval_combined_raw": 0.0396484375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1824.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2890337347984314, "rewards/retrieval_reward/mean": 1.0265625, "rewards/retrieval_reward/std": 0.28984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7173469066619873, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.23491515219211578, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.10460147857666, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20773859024047853, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1825, "step_time": 14.507798802654724 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.4546875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.2, "completions/min_terminated_length": 0.0, "entropy": 0.32003879845142363, "epoch": 7.233201581027668, "frac_reward_zero_std": 0.0, "grad_norm": 2.420785665512085, "learning_rate": 2e-05, "loss": 0.0006927490234375, "num_tokens": 109412216.0, "reward": 1.0234954833984375, "reward/completion_length_max": 35.0, "reward/completion_length_mean": 30.475, "reward/completion_length_min": 27.2, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 33.4, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.97734375, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.9265625, "reward/mean_r_ndcg": 0.8609375, "reward/mean_r_ndcg_cutoff": 6.89375, "reward/mean_recall": 0.2484375, "reward/mean_recall_at_10": 0.0490234375, "reward/mean_recall_at_100": 0.2484375, "reward/mean_retrieval_combined_bounded": 0.02430419921875, "reward/mean_retrieval_combined_raw": 0.006384063512086868, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1825.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2664422422647476, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.2669921875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6453285694122313, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.24426008462905885, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.85293197631836, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2151160269975662, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1830, "step_time": 14.639430928824003 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.42109375, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.4, "completions/min_terminated_length": 0.0, "entropy": 0.32458157613873484, "epoch": 7.252964426877471, "frac_reward_zero_std": 0.0, "grad_norm": 2.1692700386047363, "learning_rate": 2e-05, "loss": -0.0001068115234375, "num_tokens": 109707699.0, "reward": 1.01630859375, "reward/completion_length_max": 34.0, "reward/completion_length_mean": 29.4, "reward/completion_length_min": 26.4, "reward/completion_length_p90": 31.6, "reward/completion_length_p95": 32.4, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.97265625, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.89609375, "reward/mean_r_ndcg": 0.77734375, "reward/mean_r_ndcg_cutoff": 6.375, "reward/mean_recall": 0.2275390625, "reward/mean_recall_at_10": 0.04931640625, "reward/mean_recall_at_100": 0.2275390625, "reward/mean_retrieval_combined_bounded": 0.0209228515625, "reward/mean_retrieval_combined_raw": 0.024616217613220213, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1832.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2695073366165161, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.26953125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7095258235931396, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2632475733757019, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.72681999206543, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21102471947669982, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1835, "step_time": 14.51935350196436 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.875, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.4, "completions/min_terminated_length": 0.0, "entropy": 0.33363384157419207, "epoch": 7.2727272727272725, "frac_reward_zero_std": 0.0, "grad_norm": 1.8495656251907349, "learning_rate": 2e-05, "loss": 0.0005859375, "num_tokens": 110002675.0, "reward": 1.0203887939453125, "reward/completion_length_max": 33.6, "reward/completion_length_mean": 28.9, "reward/completion_length_min": 25.4, "reward/completion_length_p90": 30.8, "reward/completion_length_p95": 31.0, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.9765625, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.90703125, "reward/mean_r_ndcg": 0.8140625, "reward/mean_r_ndcg_cutoff": 6.3125, "reward/mean_recall": 0.214453125, "reward/mean_recall_at_10": 0.04814453125, "reward/mean_recall_at_100": 0.214453125, "reward/mean_retrieval_combined_bounded": 0.02236328125, "reward/mean_retrieval_combined_raw": 0.012679804721847177, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1838.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.2846346229314804, "rewards/retrieval_reward/mean": 1.0203125, "rewards/retrieval_reward/std": 0.284765625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8603065967559815, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2864948511123657, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.408482551574707, "sampling/raw_vllm_sampling_logp_difference/mean": 0.204912531375885, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1840, "step_time": 14.49253954339074 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.9515625, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.2, "completions/min_terminated_length": 0.0, "entropy": 0.35631275177001953, "epoch": 7.292490118577075, "frac_reward_zero_std": 0.0, "grad_norm": 2.090547800064087, "learning_rate": 2e-05, "loss": 0.002197265625, "num_tokens": 110297013.0, "reward": 1.0112396240234376, "reward/completion_length_max": 33.2, "reward/completion_length_mean": 28.95, "reward/completion_length_min": 26.2, "reward/completion_length_p90": 31.0, "reward/completion_length_p95": 31.4, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0109375, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.89921875, "reward/mean_r_ndcg": 0.778125, "reward/mean_r_ndcg_cutoff": 6.14375, "reward/mean_recall": 0.22734375, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.22734375, "reward/mean_retrieval_combined_bounded": 0.0152099609375, "reward/mean_retrieval_combined_raw": 0.024180126190185548, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1840.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28230518102645874, "rewards/retrieval_reward/mean": 1.0109375, "rewards/retrieval_reward/std": 0.2828125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.851697826385498, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.29980710744857786, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.253591346740723, "sampling/raw_vllm_sampling_logp_difference/mean": 0.19333203434944152, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1845, "step_time": 14.685499207640532 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.79140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.8, "completions/min_terminated_length": 0.0, "entropy": 0.3256739478558302, "epoch": 7.312252964426877, "frac_reward_zero_std": 0.0, "grad_norm": 2.256213665008545, "learning_rate": 2e-05, "loss": 0.00031890869140625, "num_tokens": 110592906.0, "reward": 1.01651611328125, "reward/completion_length_max": 33.6, "reward/completion_length_mean": 29.825, "reward/completion_length_min": 26.8, "reward/completion_length_p90": 32.4, "reward/completion_length_p95": 32.8, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.915625, "reward/mean_r_ndcg": 0.81484375, "reward/mean_r_ndcg_cutoff": 6.0, "reward/mean_recall": 0.221875, "reward/mean_recall_at_10": 0.049072265625, "reward/mean_recall_at_100": 0.221875, "reward/mean_retrieval_combined_bounded": 0.0172454833984375, "reward/mean_retrieval_combined_raw": 0.004919213056564331, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1848.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28156994581222533, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.28125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7415929794311524, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2988562613725662, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.813965606689453, "sampling/raw_vllm_sampling_logp_difference/mean": 0.18652025759220123, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1850, "step_time": 14.421431504597422 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 39.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.3046875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.25755602307617664, "epoch": 7.316205533596838, "frac_reward_zero_std": 0.0, "num_tokens": 110652216.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6775291470958354, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7302139669268998, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9086115992970123, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9973637961335676, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6191803521630042, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6755917673011682, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8666666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3247505022538783, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3622865505779089, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5277108433734939, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8481927710843373, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.621399956134714, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6365471629707604, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7661102981525517, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9151156941649901, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5738376790823969, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3654966669555709, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17963729970872397, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.31745568663390694, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 1.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3451.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.41755595247600125, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.45636759290868334, "retrieval_eval/nq/single_turn/recall_at_10": 0.6598570462667825, "retrieval_eval/nq/single_turn/recall_at_100": 0.8934849801989762, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6661581156000304, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.6960309686635736, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8360000000000001, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9486666666666668, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0262603759765625, "reward/completion_length_max": 39.0, "reward/completion_length_mean": 30.25, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0234375, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.8671875, "reward/mean_r_ndcg": 0.78125, "reward/mean_r_ndcg_cutoff": 4.625, "reward/mean_recall": 0.265625, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.265625, "reward/mean_retrieval_combined_bounded": 0.033935546875, "reward/mean_retrieval_combined_raw": 0.038330078125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1848.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2932692766189575, "rewards/retrieval_reward/mean": 1.0234375, "rewards/retrieval_reward/std": 0.29296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.950242519378662, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.4126669466495514, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 4.317789077758789, "sampling/raw_vllm_sampling_logp_difference/mean": 0.12152605503797531, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1851, "step_time": 14.173937285959255 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.75, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.45703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.75, "completions/min_terminated_length": 0.0, "entropy": 0.3048211592249572, "epoch": 7.33201581027668, "frac_reward_zero_std": 0.0, "grad_norm": 2.295971155166626, "learning_rate": 2e-05, "loss": 9.04083251953125e-05, "num_tokens": 110889228.0, "reward": 1.0131950378417969, "reward/completion_length_max": 35.75, "reward/completion_length_mean": 30.4375, "reward/completion_length_min": 26.75, "reward/completion_length_p90": 33.5, "reward/completion_length_p95": 34.25, "reward/duplicate_query_rate": 0.0009765625, "reward/duplicate_ranked_plain_query_rate": 0.0009765625, "reward/invalid_output_rate": 0.0009765625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.974609375, "reward/mean_mrr": 0.9990234375, "reward/mean_ndcg_at_10": 0.939453125, "reward/mean_r_ndcg": 0.8720703125, "reward/mean_r_ndcg_cutoff": 7.1171875, "reward/mean_recall": 0.232666015625, "reward/mean_recall_at_10": 0.04901123046875, "reward/mean_recall_at_100": 0.232666015625, "reward/mean_retrieval_combined_bounded": 0.014068603515625, "reward/mean_retrieval_combined_raw": 0.008004531264305115, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9990234375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9990234375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1854.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2825973853468895, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.283203125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.928811550140381, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.3702971190214157, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.586817860603333, "sampling/raw_vllm_sampling_logp_difference/mean": 0.17328574135899544, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1855, "step_time": 14.768555586764705 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.17421875, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.6, "completions/min_terminated_length": 0.0, "entropy": 0.33238807767629625, "epoch": 7.351778656126482, "frac_reward_zero_std": 0.0, "grad_norm": 2.3162412643432617, "learning_rate": 2e-05, "loss": 0.002001953125, "num_tokens": 111185899.0, "reward": 1.01732177734375, "reward/completion_length_max": 35.4, "reward/completion_length_mean": 30.15, "reward/completion_length_min": 26.6, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 32.6, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.8796875, "reward/mean_r_ndcg": 0.80234375, "reward/mean_r_ndcg_cutoff": 6.1, "reward/mean_recall": 0.22578125, "reward/mean_recall_at_10": 0.0486328125, "reward/mean_recall_at_100": 0.22578125, "reward/mean_retrieval_combined_bounded": 0.02149658203125, "reward/mean_retrieval_combined_raw": 0.025446224212646484, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1856.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.00078125, "reward/too_short_query_rate": 0.0, "reward_std": 0.27719916105270387, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.27734375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8333595752716065, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.30906222462654115, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.623338317871093, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20027559399604797, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1860, "step_time": 14.630859059235082 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.1109375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.33116244748234747, "epoch": 7.371541501976284, "frac_reward_zero_std": 0.0, "grad_norm": 2.2995805740356445, "learning_rate": 2e-05, "loss": -0.00218505859375, "num_tokens": 111481657.0, "reward": 1.01513671875, "reward/completion_length_max": 34.2, "reward/completion_length_mean": 30.1, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 32.4, "reward/completion_length_p95": 32.6, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.97421875, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.89375, "reward/mean_r_ndcg": 0.828125, "reward/mean_r_ndcg_cutoff": 5.8375, "reward/mean_recall": 0.2470703125, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.2470703125, "reward/mean_retrieval_combined_bounded": 0.01661376953125, "reward/mean_retrieval_combined_raw": 0.012604112178087235, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1862.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27171536684036257, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.271875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.899740219116211, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.30449526011943817, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.503523826599121, "sampling/raw_vllm_sampling_logp_difference/mean": 0.1741161435842514, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1865, "step_time": 14.56332130896626 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.31328125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.3377558641135693, "epoch": 7.391304347826087, "frac_reward_zero_std": 0.0, "grad_norm": 1.9758079051971436, "learning_rate": 2e-05, "loss": -0.001531982421875, "num_tokens": 111777770.0, "reward": 1.029022216796875, "reward/completion_length_max": 34.4, "reward/completion_length_mean": 30.3, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 32.8, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0296875, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.9203125, "reward/mean_r_ndcg": 0.83984375, "reward/mean_r_ndcg_cutoff": 6.2875, "reward/mean_recall": 0.215234375, "reward/mean_recall_at_10": 0.048876953125, "reward/mean_recall_at_100": 0.215234375, "reward/mean_retrieval_combined_bounded": 0.03040771484375, "reward/mean_retrieval_combined_raw": 0.012796003371477127, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1867.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2623285174369812, "rewards/retrieval_reward/mean": 1.0296875, "rewards/retrieval_reward/std": 0.262890625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.85801248550415, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.3263268947601318, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 5.7874753952026365, "sampling/raw_vllm_sampling_logp_difference/mean": 0.162677001953125, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1870, "step_time": 14.749567093420774 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.97734375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.8, "completions/min_terminated_length": 0.0, "entropy": 0.353409643471241, "epoch": 7.41106719367589, "frac_reward_zero_std": 0.0, "grad_norm": 2.334733486175537, "learning_rate": 2e-05, "loss": 0.001446533203125, "num_tokens": 112074893.0, "reward": 1.0086151123046876, "reward/completion_length_max": 35.8, "reward/completion_length_mean": 30.975, "reward/completion_length_min": 27.8, "reward/completion_length_p90": 33.2, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.00859375, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.9078125, "reward/mean_r_ndcg": 0.834375, "reward/mean_r_ndcg_cutoff": 6.08125, "reward/mean_recall": 0.2388671875, "reward/mean_recall_at_10": 0.04892578125, "reward/mean_recall_at_100": 0.2388671875, "reward/mean_retrieval_combined_bounded": 0.0127197265625, "reward/mean_retrieval_combined_raw": -0.0512691929936409, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1872.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2808388382196426, "rewards/retrieval_reward/mean": 1.00859375, "rewards/retrieval_reward/std": 0.2814453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7951420307159425, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.35033953189849854, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.6369068145751955, "sampling/raw_vllm_sampling_logp_difference/mean": 0.16498310565948487, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1875, "step_time": 14.693040179787204 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.47578125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.6, "completions/min_terminated_length": 0.0, "entropy": 0.3762105032801628, "epoch": 7.430830039525691, "frac_reward_zero_std": 0.0, "grad_norm": 2.0051541328430176, "learning_rate": 2e-05, "loss": 0.00026092529296875, "num_tokens": 112372750.0, "reward": 1.0135406494140624, "reward/completion_length_max": 36.2, "reward/completion_length_mean": 31.5, "reward/completion_length_min": 28.6, "reward/completion_length_p90": 34.2, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0140625, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.89375, "reward/mean_r_ndcg": 0.77578125, "reward/mean_r_ndcg_cutoff": 7.0, "reward/mean_recall": 0.2345703125, "reward/mean_recall_at_10": 0.049169921875, "reward/mean_recall_at_100": 0.2345703125, "reward/mean_retrieval_combined_bounded": 0.01654648408293724, "reward/mean_retrieval_combined_raw": 0.017130976915359496, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1876.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28644471168518065, "rewards/retrieval_reward/mean": 1.0140625, "rewards/retrieval_reward/std": 0.28671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.712139844894409, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.3121971249580383, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.324294471740723, "sampling/raw_vllm_sampling_logp_difference/mean": 0.17112701535224914, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1880, "step_time": 14.986111570533831 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.15546875, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.4, "completions/min_terminated_length": 0.0, "entropy": 0.3786665961146355, "epoch": 7.450592885375494, "frac_reward_zero_std": 0.0, "grad_norm": 2.1614573001861572, "learning_rate": 2e-05, "loss": -0.0007049560546875, "num_tokens": 112671765.0, "reward": 1.0131011962890626, "reward/completion_length_max": 36.4, "reward/completion_length_mean": 32.15, "reward/completion_length_min": 29.4, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 35.2, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.00546875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01328125, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.99453125, "reward/mean_ndcg_at_10": 0.896875, "reward/mean_r_ndcg": 0.76171875, "reward/mean_r_ndcg_cutoff": 6.1375, "reward/mean_recall": 0.2431640625, "reward/mean_recall_at_10": 0.049267578125, "reward/mean_recall_at_100": 0.2431640625, "reward/mean_retrieval_combined_bounded": 0.01856231689453125, "reward/mean_retrieval_combined_raw": 0.03193370550870896, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99453125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99453125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1881.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.290719074010849, "rewards/retrieval_reward/mean": 1.01328125, "rewards/retrieval_reward/std": 0.29140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8361114501953124, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2666969448328018, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.9601846694946286, "sampling/raw_vllm_sampling_logp_difference/mean": 0.1927517294883728, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1885, "step_time": 14.805418148613535 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.5734375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.2, "completions/min_terminated_length": 0.0, "entropy": 0.38557268679142, "epoch": 7.470355731225297, "frac_reward_zero_std": 0.0, "grad_norm": 2.1081743240356445, "learning_rate": 2e-05, "loss": -0.001708984375, "num_tokens": 112970995.0, "reward": 1.044952392578125, "reward/completion_length_max": 36.6, "reward/completion_length_mean": 32.625, "reward/completion_length_min": 27.2, "reward/completion_length_p90": 34.8, "reward/completion_length_p95": 35.6, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.00546875, "reward/length_out_of_range_rate": 0.00078125, "reward/low_length_query_rate": 0.00078125, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.04375, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99453125, "reward/mean_ndcg_at_10": 0.871875, "reward/mean_r_ndcg": 0.72578125, "reward/mean_r_ndcg_cutoff": 6.5375, "reward/mean_recall": 0.2224609375, "reward/mean_recall_at_10": 0.04833984375, "reward/mean_recall_at_100": 0.2224609375, "reward/mean_retrieval_combined_bounded": 0.05068359375, "reward/mean_retrieval_combined_raw": 0.18271484375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99453125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99453125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1888.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00078125, "reward_std": 0.2859564036130905, "rewards/retrieval_reward/mean": 1.04375, "rewards/retrieval_reward/std": 0.2869140625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6556175708770753, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2761176109313965, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.873389625549317, "sampling/raw_vllm_sampling_logp_difference/mean": 0.1988297790288925, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1890, "step_time": 14.970002071012278 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.39296875, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.3899298973381519, "epoch": 7.490118577075099, "frac_reward_zero_std": 0.0, "grad_norm": 2.1719722747802734, "learning_rate": 2e-05, "loss": -0.000506591796875, "num_tokens": 113269834.0, "reward": 1.016485595703125, "reward/completion_length_max": 36.8, "reward/completion_length_mean": 32.4, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 34.6, "reward/completion_length_p95": 35.0, "reward/duplicate_query_rate": 0.0109375, "reward/duplicate_ranked_plain_query_rate": 0.0109375, "reward/invalid_output_rate": 0.0109375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0171875, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.9890625, "reward/mean_ndcg_at_10": 0.90078125, "reward/mean_r_ndcg": 0.7453125, "reward/mean_r_ndcg_cutoff": 6.65, "reward/mean_recall": 0.266015625, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.266015625, "reward/mean_retrieval_combined_bounded": 0.027587890625, "reward/mean_retrieval_combined_raw": 0.04375176532194018, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9890625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9890625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1891.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2892078340053558, "rewards/retrieval_reward/mean": 1.0171875, "rewards/retrieval_reward/std": 0.289453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.9070555210113525, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2802184998989105, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.656782531738282, "sampling/raw_vllm_sampling_logp_difference/mean": 0.19849203824996947, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1895, "step_time": 14.849721156188753 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 32.1125, "completions/mean_terminated_length": 0.0, "completions/min_length": 29.0, "completions/min_terminated_length": 0.0, "entropy": 0.3644831791520119, "epoch": 7.509881422924901, "frac_reward_zero_std": 0.0, "grad_norm": 2.4003829956054688, "learning_rate": 2e-05, "loss": 7.74383544921875e-05, "num_tokens": 113568762.0, "reward": 1.0111572265625, "reward/completion_length_max": 36.4, "reward/completion_length_mean": 32.1, "reward/completion_length_min": 29.0, "reward/completion_length_p90": 34.4, "reward/completion_length_p95": 34.8, "reward/duplicate_query_rate": 0.025, "reward/duplicate_ranked_plain_query_rate": 0.025, "reward/invalid_output_rate": 0.025, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01171875, "reward/mean_colbert": 0.95234375, "reward/mean_mrr": 0.975, "reward/mean_ndcg_at_10": 0.89765625, "reward/mean_r_ndcg": 0.83046875, "reward/mean_r_ndcg_cutoff": 6.43125, "reward/mean_recall": 0.2259765625, "reward/mean_recall_at_10": 0.0474609375, "reward/mean_recall_at_100": 0.2259765625, "reward/mean_retrieval_combined_bounded": 0.0362548828125, "reward/mean_retrieval_combined_raw": 0.070068359375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.975, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.975, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1896.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.3137643218040466, "rewards/retrieval_reward/mean": 1.01171875, "rewards/retrieval_reward/std": 0.314453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.650082588195801, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.24667948484420776, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.167959403991699, "sampling/raw_vllm_sampling_logp_difference/mean": 0.1977691113948822, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1900, "step_time": 14.778553724999075 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.1875, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.35086773335933685, "epoch": 7.513833992094861, "frac_reward_zero_std": 0.0, "num_tokens": 113628170.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6717014087290294, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.7255571951489399, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9077328646748682, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9938488576449912, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.620917445082457, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6694208014869306, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.84, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9866666666666667, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.3397249494922576, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.37304120397594054, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5253012048192771, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8650602409638555, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.627242674600548, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.635343461338181, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7556391378574477, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.9069835680751173, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5721420295834123, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.3591461647826174, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17297757931689672, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.30779125068538404, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 1.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3451.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.42359123793727094, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4608393847921162, "retrieval_eval/nq/single_turn/recall_at_10": 0.6606297691490389, "retrieval_eval/nq/single_turn/recall_at_100": 0.8879551820728292, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6670719849193009, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7025802435793005, "retrieval_eval/scifact/single_turn/recall_at_10": 0.8456666666666667, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9553333333333334, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.017425537109375, "reward/completion_length_max": 34.0, "reward/completion_length_mean": 31.25, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 33.0, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.953125, "reward/mean_r_ndcg": 0.8359375, "reward/mean_r_ndcg_cutoff": 6.9375, "reward/mean_recall": 0.1875, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.1875, "reward/mean_retrieval_combined_bounded": 0.025146484375, "reward/mean_retrieval_combined_raw": 0.04443359375, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1904.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2746373414993286, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.275390625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8108584880828857, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2498498111963272, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.863283157348633, "sampling/raw_vllm_sampling_logp_difference/mean": 0.19714348018169403, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1901, "step_time": 14.575454011093825 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 31.3564453125, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.25, "completions/min_terminated_length": 0.0, "entropy": 0.39165288396179676, "epoch": 7.529644268774703, "frac_reward_zero_std": 0.0, "grad_norm": 2.246381998062134, "learning_rate": 2e-05, "loss": -0.00206298828125, "num_tokens": 113866455.0, "reward": 1.0128707885742188, "reward/completion_length_max": 36.0, "reward/completion_length_mean": 31.375, "reward/completion_length_min": 28.25, "reward/completion_length_p90": 33.5, "reward/completion_length_p95": 33.75, "reward/duplicate_query_rate": 0.001953125, "reward/duplicate_ranked_plain_query_rate": 0.001953125, "reward/invalid_output_rate": 0.001953125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01171875, "reward/mean_colbert": 0.9716796875, "reward/mean_mrr": 0.998046875, "reward/mean_ndcg_at_10": 0.9091796875, "reward/mean_r_ndcg": 0.767578125, "reward/mean_r_ndcg_cutoff": 6.2890625, "reward/mean_recall": 0.222412109375, "reward/mean_recall_at_10": 0.04840087890625, "reward/mean_recall_at_100": 0.222412109375, "reward/mean_retrieval_combined_bounded": 0.014942169189453125, "reward/mean_retrieval_combined_raw": 0.010314481798559427, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.998046875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.998046875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1904.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2684922590851784, "rewards/retrieval_reward/mean": 1.01171875, "rewards/retrieval_reward/std": 0.2685546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.777296483516693, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2718757428228855, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.492493391036987, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21467109397053719, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1905, "step_time": 14.693410232794122 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.91953125, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.4058696605265141, "epoch": 7.549407114624506, "frac_reward_zero_std": 0.0, "grad_norm": 2.0046699047088623, "learning_rate": 2e-05, "loss": 0.00022430419921875, "num_tokens": 114163600.0, "reward": 1.021807861328125, "reward/completion_length_max": 35.6, "reward/completion_length_mean": 30.925, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 32.8, "reward/completion_length_p95": 33.2, "reward/duplicate_query_rate": 0.0046875, "reward/duplicate_ranked_plain_query_rate": 0.0046875, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.87890625, "reward/mean_r_ndcg": 0.70859375, "reward/mean_r_ndcg_cutoff": 5.51875, "reward/mean_recall": 0.241015625, "reward/mean_recall_at_10": 0.0486328125, "reward/mean_recall_at_100": 0.241015625, "reward/mean_retrieval_combined_bounded": 0.0264251708984375, "reward/mean_retrieval_combined_raw": 0.028366422932595014, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1905.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27358673214912416, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.27421875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.590213584899902, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2646937668323517, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.178312397003173, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2058152288198471, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1910, "step_time": 14.51358085158281 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.13984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.3420948937535286, "epoch": 7.569169960474309, "frac_reward_zero_std": 0.0, "grad_norm": 2.183732509613037, "learning_rate": 2e-05, "loss": -0.001953125, "num_tokens": 114459971.0, "reward": 1.023175048828125, "reward/completion_length_max": 34.0, "reward/completion_length_mean": 30.15, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 31.4, "reward/completion_length_p95": 32.4, "reward/duplicate_query_rate": 0.00703125, "reward/duplicate_ranked_plain_query_rate": 0.00703125, "reward/invalid_output_rate": 0.00703125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.02421875, "reward/mean_colbert": 0.96953125, "reward/mean_mrr": 0.99296875, "reward/mean_ndcg_at_10": 0.9109375, "reward/mean_r_ndcg": 0.8265625, "reward/mean_r_ndcg_cutoff": 7.26875, "reward/mean_recall": 0.228125, "reward/mean_recall_at_10": 0.04931640625, "reward/mean_recall_at_100": 0.228125, "reward/mean_retrieval_combined_bounded": 0.0302947998046875, "reward/mean_retrieval_combined_raw": 0.03486095666885376, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99296875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99296875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1912.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.287897002696991, "rewards/retrieval_reward/mean": 1.02421875, "rewards/retrieval_reward/std": 0.288671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.4773838996887205, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2404440701007843, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.561453437805175, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20709118247032166, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1915, "step_time": 14.650391404586845 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.96640625, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.0, "completions/min_terminated_length": 0.0, "entropy": 0.32969155758619306, "epoch": 7.58893280632411, "frac_reward_zero_std": 0.0, "grad_norm": 2.525904417037964, "learning_rate": 2e-05, "loss": -0.0005523681640625, "num_tokens": 114755896.0, "reward": 1.026025390625, "reward/completion_length_max": 34.0, "reward/completion_length_mean": 29.95, "reward/completion_length_min": 27.0, "reward/completion_length_p90": 32.4, "reward/completion_length_p95": 32.8, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.025, "reward/mean_colbert": 0.9765625, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.89375, "reward/mean_r_ndcg": 0.75, "reward/mean_r_ndcg_cutoff": 6.225, "reward/mean_recall": 0.246875, "reward/mean_recall_at_10": 0.04853515625, "reward/mean_recall_at_100": 0.246875, "reward/mean_retrieval_combined_bounded": 0.02606201171875, "reward/mean_retrieval_combined_raw": -3.9517879486083984e-05, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1918.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.26982419192790985, "rewards/retrieval_reward/mean": 1.025, "rewards/retrieval_reward/std": 0.2703125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8613287448883056, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.30642856657505035, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.552840614318848, "sampling/raw_vllm_sampling_logp_difference/mean": 0.18271915912628173, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1920, "step_time": 14.458086413587443 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.5171875, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.8, "completions/min_terminated_length": 0.0, "entropy": 0.35279101356863973, "epoch": 7.608695652173913, "frac_reward_zero_std": 0.0, "grad_norm": 2.4756481647491455, "learning_rate": 2e-05, "loss": -0.000115203857421875, "num_tokens": 115050894.0, "reward": 1.02059326171875, "reward/completion_length_max": 33.8, "reward/completion_length_mean": 29.5, "reward/completion_length_min": 26.8, "reward/completion_length_p90": 31.6, "reward/completion_length_p95": 32.2, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.97578125, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.91171875, "reward/mean_r_ndcg": 0.8078125, "reward/mean_r_ndcg_cutoff": 5.975, "reward/mean_recall": 0.255078125, "reward/mean_recall_at_10": 0.0490234375, "reward/mean_recall_at_100": 0.255078125, "reward/mean_retrieval_combined_bounded": 0.021514892578125, "reward/mean_retrieval_combined_raw": 0.0062464654445648195, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1920.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2656367182731628, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.265625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8713626861572266, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.26898258924484253, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.17322063446045, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21378493010997773, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1925, "step_time": 14.299526809179223 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.75859375, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.2, "completions/min_terminated_length": 0.0, "entropy": 0.33983128182590006, "epoch": 7.628458498023716, "frac_reward_zero_std": 0.0, "grad_norm": 2.1529576778411865, "learning_rate": 2e-05, "loss": -0.00257568359375, "num_tokens": 115345433.0, "reward": 1.0184539794921874, "reward/completion_length_max": 32.2, "reward/completion_length_mean": 28.75, "reward/completion_length_min": 25.2, "reward/completion_length_p90": 30.8, "reward/completion_length_p95": 31.4, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.0046875, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01796875, "reward/mean_colbert": 0.971875, "reward/mean_mrr": 0.9953125, "reward/mean_ndcg_at_10": 0.9171875, "reward/mean_r_ndcg": 0.8015625, "reward/mean_r_ndcg_cutoff": 6.06875, "reward/mean_recall": 0.2248046875, "reward/mean_recall_at_10": 0.04921875, "reward/mean_recall_at_100": 0.2248046875, "reward/mean_retrieval_combined_bounded": 0.02411346435546875, "reward/mean_retrieval_combined_raw": 0.029715202376246454, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9953125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9953125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1928.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0015625, "reward_std": 0.28578892946243284, "rewards/retrieval_reward/mean": 1.01796875, "rewards/retrieval_reward/std": 0.28671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.723737382888794, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2908130526542664, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.755425548553466, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20323496758937837, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1930, "step_time": 14.640554260381032 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.3703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.6, "completions/min_terminated_length": 0.0, "entropy": 0.3142180770635605, "epoch": 7.648221343873518, "frac_reward_zero_std": 0.0, "grad_norm": 2.5203633308410645, "learning_rate": 2e-05, "loss": -0.000634765625, "num_tokens": 115639091.0, "reward": 1.0117919921875, "reward/completion_length_max": 32.4, "reward/completion_length_mean": 28.375, "reward/completion_length_min": 25.6, "reward/completion_length_p90": 30.2, "reward/completion_length_p95": 30.6, "reward/duplicate_query_rate": 0.00078125, "reward/duplicate_ranked_plain_query_rate": 0.00078125, "reward/invalid_output_rate": 0.00078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01171875, "reward/mean_colbert": 0.975, "reward/mean_mrr": 0.99921875, "reward/mean_ndcg_at_10": 0.91328125, "reward/mean_r_ndcg": 0.76328125, "reward/mean_r_ndcg_cutoff": 6.01875, "reward/mean_recall": 0.246875, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.246875, "reward/mean_retrieval_combined_bounded": 0.0125701904296875, "reward/mean_retrieval_combined_raw": 0.005676903203129769, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1932.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.26248501539230346, "rewards/retrieval_reward/mean": 1.01171875, "rewards/retrieval_reward/std": 0.2630859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.822275972366333, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.34171451926231383, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.007170104980469, "sampling/raw_vllm_sampling_logp_difference/mean": 0.196152862906456, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1935, "step_time": 14.435926544794347 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 31.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 27.2890625, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.4, "completions/min_terminated_length": 0.0, "entropy": 0.34160138443112376, "epoch": 7.66798418972332, "frac_reward_zero_std": 0.0, "grad_norm": 2.9008402824401855, "learning_rate": 2e-05, "loss": -0.00084228515625, "num_tokens": 115931077.0, "reward": 1.01474609375, "reward/completion_length_max": 31.6, "reward/completion_length_mean": 27.3, "reward/completion_length_min": 24.4, "reward/completion_length_p90": 28.8, "reward/completion_length_p95": 29.2, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.009375, "reward/length_out_of_range_rate": 0.009375, "reward/low_length_query_rate": 0.009375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.96640625, "reward/mean_mrr": 0.990625, "reward/mean_ndcg_at_10": 0.91796875, "reward/mean_r_ndcg": 0.77890625, "reward/mean_r_ndcg_cutoff": 6.65625, "reward/mean_recall": 0.234375, "reward/mean_recall_at_10": 0.04912109375, "reward/mean_recall_at_100": 0.234375, "reward/mean_retrieval_combined_bounded": 0.02880859375, "reward/mean_retrieval_combined_raw": 0.06543493270874023, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.990625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.990625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1936.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.009375, "reward_std": 0.3263252705335617, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.3265625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.773523187637329, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2844809293746948, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.580113792419434, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2028382033109665, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1940, "step_time": 13.921683738438878 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 31.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 27.0765625, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.4, "completions/min_terminated_length": 0.0, "entropy": 0.2977480415254831, "epoch": 7.687747035573123, "frac_reward_zero_std": 0.0, "grad_norm": 2.3382112979888916, "learning_rate": 2e-05, "loss": -0.0008056640625, "num_tokens": 116223079.0, "reward": 1.0035003662109374, "reward/completion_length_max": 31.2, "reward/completion_length_mean": 27.075, "reward/completion_length_min": 24.4, "reward/completion_length_p90": 29.2, "reward/completion_length_p95": 29.6, "reward/duplicate_query_rate": 0.00546875, "reward/duplicate_ranked_plain_query_rate": 0.00546875, "reward/invalid_output_rate": 0.01171875, "reward/length_out_of_range_rate": 0.00625, "reward/low_length_query_rate": 0.00625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.003125, "reward/mean_colbert": 0.9625, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.859375, "reward/mean_r_ndcg": 0.72734375, "reward/mean_r_ndcg_cutoff": 5.80625, "reward/mean_recall": 0.2427734375, "reward/mean_recall_at_10": 0.047802734375, "reward/mean_recall_at_100": 0.2427734375, "reward/mean_retrieval_combined_bounded": 0.018274307250976562, "reward/mean_retrieval_combined_raw": 0.03534998595714569, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1942.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00625, "reward_std": 0.3102979302406311, "rewards/retrieval_reward/mean": 1.0046875, "rewards/retrieval_reward/std": 0.310546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8298989295959474, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.34152293801307676, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.134233665466308, "sampling/raw_vllm_sampling_logp_difference/mean": 0.21177177727222443, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1945, "step_time": 14.13322302675806 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 30.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 27.5703125, "completions/mean_terminated_length": 0.0, "completions/min_length": 24.4, "completions/min_terminated_length": 0.0, "entropy": 0.30492656081914904, "epoch": 7.707509881422925, "frac_reward_zero_std": 0.0, "grad_norm": 2.611248731613159, "learning_rate": 2e-05, "loss": -0.00274658203125, "num_tokens": 116515841.0, "reward": 1.0211578369140626, "reward/completion_length_max": 30.8, "reward/completion_length_mean": 27.6, "reward/completion_length_min": 24.4, "reward/completion_length_p90": 29.4, "reward/completion_length_p95": 29.8, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.00859375, "reward/length_out_of_range_rate": 0.00859375, "reward/low_length_query_rate": 0.00859375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.99140625, "reward/mean_ndcg_at_10": 0.8953125, "reward/mean_r_ndcg": 0.7953125, "reward/mean_r_ndcg_cutoff": 6.75625, "reward/mean_recall": 0.2318359375, "reward/mean_recall_at_10": 0.048193359375, "reward/mean_recall_at_100": 0.2318359375, "reward/mean_retrieval_combined_bounded": 0.0341796875, "reward/mean_retrieval_combined_raw": 0.03080894947052002, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99140625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99140625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1947.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.00859375, "reward_std": 0.29374096393585203, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.29453125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.869563579559326, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.3585571885108948, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 6.2458909034729, "sampling/raw_vllm_sampling_logp_difference/mean": 0.18077943325042725, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1950, "step_time": 14.388403763528913 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 32.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 27.94140625, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.0, "completions/min_terminated_length": 0.0, "entropy": 0.322091992944479, "epoch": 7.711462450592886, "frac_reward_zero_std": 0.0, "num_tokens": 116574386.0, "retrieval_eval/FinQARetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinQARetrieval/single_turn/evaluated_queries": 1138.0, "retrieval_eval/FinQARetrieval/single_turn/mrr_at_100": 0.6765862422469004, "retrieval_eval/FinQARetrieval/single_turn/ndcg_at_10": 0.730061864658471, "retrieval_eval/FinQARetrieval/single_turn/recall_at_10": 0.9103690685413005, "retrieval_eval/FinQARetrieval/single_turn/recall_at_100": 0.9956063268892794, "retrieval_eval/FinQARetrieval/single_turn/total_queries": 1138.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/evaluated_queries": 150.0, "retrieval_eval/FinanceBenchRetrieval/single_turn/mrr_at_100": 0.6197877233534894, "retrieval_eval/FinanceBenchRetrieval/single_turn/ndcg_at_10": 0.6663841193713682, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_10": 0.8333333333333334, "retrieval_eval/FinanceBenchRetrieval/single_turn/recall_at_100": 0.9933333333333333, "retrieval_eval/FinanceBenchRetrieval/single_turn/total_queries": 150.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/dropped_queries": 0.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/evaluated_queries": 415.0, "retrieval_eval/HC3FinanceRetrieval/single_turn/mrr_at_100": 0.33269277611751863, "retrieval_eval/HC3FinanceRetrieval/single_turn/ndcg_at_10": 0.3720878863592874, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_10": 0.5421686746987951, "retrieval_eval/HC3FinanceRetrieval/single_turn/recall_at_100": 0.8602409638554217, "retrieval_eval/HC3FinanceRetrieval/single_turn/total_queries": 415.0, "retrieval_eval/LegalSummarization/single_turn/dropped_queries": 0.0, "retrieval_eval/LegalSummarization/single_turn/evaluated_queries": 284.0, "retrieval_eval/LegalSummarization/single_turn/mrr_at_100": 0.6278873966682473, "retrieval_eval/LegalSummarization/single_turn/ndcg_at_10": 0.6384608118292636, "retrieval_eval/LegalSummarization/single_turn/recall_at_10": 0.7661061063349796, "retrieval_eval/LegalSummarization/single_turn/recall_at_100": 0.91030348759222, "retrieval_eval/LegalSummarization/single_turn/total_queries": 284.0, "retrieval_eval/nfcorpus/single_turn/dropped_queries": 0.0, "retrieval_eval/nfcorpus/single_turn/evaluated_queries": 323.0, "retrieval_eval/nfcorpus/single_turn/mrr_at_100": 0.5628884408139787, "retrieval_eval/nfcorpus/single_turn/ndcg_at_10": 0.35895618722496536, "retrieval_eval/nfcorpus/single_turn/recall_at_10": 0.17772470173289834, "retrieval_eval/nfcorpus/single_turn/recall_at_100": 0.3112792146218096, "retrieval_eval/nfcorpus/single_turn/total_queries": 323.0, "retrieval_eval/nq/single_turn/dropped_queries": 1.0, "retrieval_eval/nq/single_turn/evaluated_queries": 3451.0, "retrieval_eval/nq/single_turn/mrr_at_100": 0.41179304382671067, "retrieval_eval/nq/single_turn/ndcg_at_10": 0.4485009191904664, "retrieval_eval/nq/single_turn/recall_at_10": 0.6477349560513861, "retrieval_eval/nq/single_turn/recall_at_100": 0.8845745194629575, "retrieval_eval/nq/single_turn/total_queries": 3452.0, "retrieval_eval/scifact/single_turn/dropped_queries": 0.0, "retrieval_eval/scifact/single_turn/evaluated_queries": 300.0, "retrieval_eval/scifact/single_turn/mrr_at_100": 0.6829945948947903, "retrieval_eval/scifact/single_turn/ndcg_at_10": 0.7100982400302238, "retrieval_eval/scifact/single_turn/recall_at_10": 0.835, "retrieval_eval/scifact/single_turn/recall_at_100": 0.9582222222222221, "retrieval_eval/scifact/single_turn/total_queries": 300.0, "reward": 1.0415802001953125, "reward/completion_length_max": 32.0, "reward/completion_length_mean": 28.0, "reward/completion_length_min": 26.0, "reward/completion_length_p90": 30.0, "reward/completion_length_p95": 30.0, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0390625, "reward/mean_colbert": 0.9765625, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.87890625, "reward/mean_r_ndcg": 0.7890625, "reward/mean_r_ndcg_cutoff": 5.375, "reward/mean_recall": 0.232421875, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.232421875, "reward/mean_retrieval_combined_bounded": 0.045654296875, "reward/mean_retrieval_combined_raw": 0.03173828125, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1952.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2631804347038269, "rewards/retrieval_reward/mean": 1.0390625, "rewards/retrieval_reward/std": 0.263671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8402247428894043, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.3864535093307495, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.59715461730957, "sampling/raw_vllm_sampling_logp_difference/mean": 0.16411639750003815, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1951, "step_time": 14.456625936087221 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 31.25, "completions/max_terminated_length": 0.0, "completions/mean_length": 27.697265625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.0, "completions/min_terminated_length": 0.0, "entropy": 0.32299792394042015, "epoch": 7.7272727272727275, "frac_reward_zero_std": 0.0, "grad_norm": 2.291621446609497, "learning_rate": 2e-05, "loss": -0.00244140625, "num_tokens": 116808604.0, "reward": 1.0024757385253906, "reward/completion_length_max": 31.25, "reward/completion_length_mean": 27.71875, "reward/completion_length_min": 25.0, "reward/completion_length_p90": 29.5, "reward/completion_length_p95": 30.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.005859375, "reward/length_out_of_range_rate": 0.005859375, "reward/low_length_query_rate": 0.005859375, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.001953125, "reward/mean_colbert": 0.96875, "reward/mean_mrr": 0.994140625, "reward/mean_ndcg_at_10": 0.87890625, "reward/mean_r_ndcg": 0.7216796875, "reward/mean_r_ndcg_cutoff": 5.75, "reward/mean_recall": 0.231201171875, "reward/mean_recall_at_10": 0.048583984375, "reward/mean_recall_at_100": 0.231201171875, "reward/mean_retrieval_combined_bounded": 0.011357784271240234, "reward/mean_retrieval_combined_raw": 0.017687231302261353, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.994140625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.994140625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1952.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.005859375, "reward_std": 0.3113136291503906, "rewards/retrieval_reward/mean": 1.001953125, "rewards/retrieval_reward/std": 0.31201171875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.734305679798126, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.32752713561058044, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.123523712158203, "sampling/raw_vllm_sampling_logp_difference/mean": 0.18479850143194199, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1955, "step_time": 13.535029632999795 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 28.2390625, "completions/mean_terminated_length": 0.0, "completions/min_length": 25.4, "completions/min_terminated_length": 0.0, "entropy": 0.3285907916724682, "epoch": 7.747035573122529, "frac_reward_zero_std": 0.0, "grad_norm": 2.242786407470703, "learning_rate": 2e-05, "loss": 7.51495361328125e-05, "num_tokens": 117102030.0, "reward": 1.0301300048828126, "reward/completion_length_max": 33.0, "reward/completion_length_mean": 28.225, "reward/completion_length_min": 25.4, "reward/completion_length_p90": 30.6, "reward/completion_length_p95": 31.0, "reward/duplicate_query_rate": 0.01015625, "reward/duplicate_ranked_plain_query_rate": 0.01015625, "reward/invalid_output_rate": 0.01171875, "reward/length_out_of_range_rate": 0.0015625, "reward/low_length_query_rate": 0.0015625, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.03125, "reward/mean_colbert": 0.96484375, "reward/mean_mrr": 0.98828125, "reward/mean_ndcg_at_10": 0.91484375, "reward/mean_r_ndcg": 0.81328125, "reward/mean_r_ndcg_cutoff": 5.775, "reward/mean_recall": 0.2466796875, "reward/mean_recall_at_10": 0.04853515625, "reward/mean_recall_at_100": 0.2466796875, "reward/mean_retrieval_combined_bounded": 0.0426513671875, "reward/mean_retrieval_combined_raw": 0.1078033596277237, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.98828125, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.98828125, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1956.8, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0015625, "reward_std": 0.29347090125083924, "rewards/retrieval_reward/mean": 1.0296875, "rewards/retrieval_reward/std": 0.2935546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7435425758361816, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.3133935213088989, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 7.824469375610351, "sampling/raw_vllm_sampling_logp_difference/mean": 0.1801505833864212, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1960, "step_time": 13.565220745559781 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 33.4, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.11484375, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.0, "completions/min_terminated_length": 0.0, "entropy": 0.3237056747078896, "epoch": 7.766798418972332, "frac_reward_zero_std": 0.0, "grad_norm": 2.2388479709625244, "learning_rate": 2e-05, "loss": 0.0005767822265625, "num_tokens": 117396801.0, "reward": 1.0216217041015625, "reward/completion_length_max": 33.4, "reward/completion_length_mean": 29.1, "reward/completion_length_min": 26.0, "reward/completion_length_p90": 31.6, "reward/completion_length_p95": 32.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.021875, "reward/mean_colbert": 0.975, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.90859375, "reward/mean_r_ndcg": 0.76484375, "reward/mean_r_ndcg_cutoff": 5.875, "reward/mean_recall": 0.23671875, "reward/mean_recall_at_10": 0.04921875, "reward/mean_recall_at_100": 0.23671875, "reward/mean_retrieval_combined_bounded": 0.0216888427734375, "reward/mean_retrieval_combined_raw": 0.07460813913494349, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1961.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27107688784599304, "rewards/retrieval_reward/mean": 1.021875, "rewards/retrieval_reward/std": 0.271484375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.823276996612549, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.28845040798187255, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.585340118408203, "sampling/raw_vllm_sampling_logp_difference/mean": 0.18061905205249787, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1965, "step_time": 13.657097531401087 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 29.7546875, "completions/mean_terminated_length": 0.0, "completions/min_length": 26.8, "completions/min_terminated_length": 0.0, "entropy": 0.339597561955452, "epoch": 7.786561264822135, "frac_reward_zero_std": 0.0, "grad_norm": 2.0660400390625, "learning_rate": 2e-05, "loss": 0.001470947265625, "num_tokens": 117692231.0, "reward": 1.016986083984375, "reward/completion_length_max": 34.0, "reward/completion_length_mean": 29.75, "reward/completion_length_min": 26.8, "reward/completion_length_p90": 32.0, "reward/completion_length_p95": 32.4, "reward/duplicate_query_rate": 0.00390625, "reward/duplicate_ranked_plain_query_rate": 0.00390625, "reward/invalid_output_rate": 0.00390625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 0.99609375, "reward/mean_ndcg_at_10": 0.91875, "reward/mean_r_ndcg": 0.7890625, "reward/mean_r_ndcg_cutoff": 6.95, "reward/mean_recall": 0.2515625, "reward/mean_recall_at_10": 0.049072265625, "reward/mean_recall_at_100": 0.2515625, "reward/mean_retrieval_combined_bounded": 0.02086181640625, "reward/mean_retrieval_combined_raw": 0.01855285167694092, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99609375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99609375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1968.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2830700635910034, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.283984375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.6754081726074217, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.27227037847042085, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.812356948852539, "sampling/raw_vllm_sampling_logp_difference/mean": 0.2028593510389328, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1970, "step_time": 13.66186559612397 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.4546875, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.34520498514175413, "epoch": 7.8063241106719365, "frac_reward_zero_std": 0.0, "grad_norm": 2.406745672225952, "learning_rate": 2e-05, "loss": -0.00142822265625, "num_tokens": 117988781.0, "reward": 1.0156097412109375, "reward/completion_length_max": 34.8, "reward/completion_length_mean": 30.45, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 31.8, "reward/completion_length_p95": 32.2, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.015625, "reward/mean_colbert": 0.9734375, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.8875, "reward/mean_r_ndcg": 0.7796875, "reward/mean_r_ndcg_cutoff": 5.6, "reward/mean_recall": 0.2171875, "reward/mean_recall_at_10": 0.048681640625, "reward/mean_recall_at_100": 0.2171875, "reward/mean_retrieval_combined_bounded": 0.0156982421875, "reward/mean_retrieval_combined_raw": 4.1128695011138915e-05, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1971.2, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27903730869293214, "rewards/retrieval_reward/mean": 1.015625, "rewards/retrieval_reward/std": 0.279296875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.7010776519775392, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.28061923384666443, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.854967308044433, "sampling/raw_vllm_sampling_logp_difference/mean": 0.1918931484222412, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1975, "step_time": 13.708260141173378 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.8, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.6984375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.3313869141042233, "epoch": 7.826086956521739, "frac_reward_zero_std": 0.0, "grad_norm": 2.7979328632354736, "learning_rate": 2e-05, "loss": -0.000811767578125, "num_tokens": 118285867.0, "reward": 1.0419830322265624, "reward/completion_length_max": 34.8, "reward/completion_length_mean": 30.7, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.0015625, "reward/duplicate_ranked_plain_query_rate": 0.0015625, "reward/invalid_output_rate": 0.0015625, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0421875, "reward/mean_colbert": 0.97421875, "reward/mean_mrr": 0.9984375, "reward/mean_ndcg_at_10": 0.9171875, "reward/mean_r_ndcg": 0.77421875, "reward/mean_r_ndcg_cutoff": 6.5875, "reward/mean_recall": 0.25625, "reward/mean_recall_at_10": 0.04912109375, "reward/mean_recall_at_100": 0.25625, "reward/mean_retrieval_combined_bounded": 0.0434814453125, "reward/mean_retrieval_combined_raw": 0.15820188000798224, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9984375, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9984375, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1976.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2854245722293854, "rewards/retrieval_reward/mean": 1.0421875, "rewards/retrieval_reward/std": 0.2859375, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.812789249420166, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.24563279747962952, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.628745460510254, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22070254385471344, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1980, "step_time": 13.800254663999658 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 36.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.67109375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.3825149968266487, "epoch": 7.845849802371541, "frac_reward_zero_std": 0.0, "grad_norm": 2.08845853805542, "learning_rate": 2e-05, "loss": 0.0006103515625, "num_tokens": 118582502.0, "reward": 1.0115997314453125, "reward/completion_length_max": 36.2, "reward/completion_length_mean": 30.675, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 32.6, "reward/completion_length_p95": 33.0, "reward/duplicate_query_rate": 0.0078125, "reward/duplicate_ranked_plain_query_rate": 0.0078125, "reward/invalid_output_rate": 0.0078125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.0125, "reward/mean_colbert": 0.96796875, "reward/mean_mrr": 0.9921875, "reward/mean_ndcg_at_10": 0.91015625, "reward/mean_r_ndcg": 0.7546875, "reward/mean_r_ndcg_cutoff": 6.78125, "reward/mean_recall": 0.22109375, "reward/mean_recall_at_10": 0.048828125, "reward/mean_recall_at_100": 0.22109375, "reward/mean_retrieval_combined_bounded": 0.01944427490234375, "reward/mean_retrieval_combined_raw": 0.03304597437381744, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.9921875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.9921875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1984.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.28566539883613584, "rewards/retrieval_reward/mean": 1.0125, "rewards/retrieval_reward/std": 0.285546875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8429996013641357, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.28118720948696135, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 10.570546913146973, "sampling/raw_vllm_sampling_logp_difference/mean": 0.19682693481445312, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1985, "step_time": 13.804281044425442 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.2, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.54609375, "completions/mean_terminated_length": 0.0, "completions/min_length": 27.6, "completions/min_terminated_length": 0.0, "entropy": 0.3626104071736336, "epoch": 7.865612648221344, "frac_reward_zero_std": 0.0, "grad_norm": 2.7419545650482178, "learning_rate": 2e-05, "loss": -0.00016021728515625, "num_tokens": 118879169.0, "reward": 1.0173065185546875, "reward/completion_length_max": 35.2, "reward/completion_length_mean": 30.525, "reward/completion_length_min": 27.6, "reward/completion_length_p90": 32.2, "reward/completion_length_p95": 32.8, "reward/duplicate_query_rate": 0.003125, "reward/duplicate_ranked_plain_query_rate": 0.003125, "reward/invalid_output_rate": 0.003125, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01796875, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.996875, "reward/mean_ndcg_at_10": 0.88828125, "reward/mean_r_ndcg": 0.76328125, "reward/mean_r_ndcg_cutoff": 5.4375, "reward/mean_recall": 0.2333984375, "reward/mean_recall_at_10": 0.049072265625, "reward/mean_recall_at_100": 0.2333984375, "reward/mean_retrieval_combined_bounded": 0.020501708984375, "reward/mean_retrieval_combined_raw": 0.02133329873904586, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.996875, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.996875, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1985.6, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.27785252332687377, "rewards/retrieval_reward/mean": 1.01796875, "rewards/retrieval_reward/std": 0.278125, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.893618440628052, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.2568595826625824, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 9.315171432495116, "sampling/raw_vllm_sampling_logp_difference/mean": 0.22447350025177001, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1990, "step_time": 13.808575157425366 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 34.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.7109375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.0, "completions/min_terminated_length": 0.0, "entropy": 0.36226756274700167, "epoch": 7.8853754940711465, "frac_reward_zero_std": 0.0, "grad_norm": 2.417043447494507, "learning_rate": 2e-05, "loss": -0.001556396484375, "num_tokens": 119176047.0, "reward": 1.0267974853515625, "reward/completion_length_max": 34.6, "reward/completion_length_mean": 30.7, "reward/completion_length_min": 28.0, "reward/completion_length_p90": 32.2, "reward/completion_length_p95": 32.6, "reward/duplicate_query_rate": 0.00234375, "reward/duplicate_ranked_plain_query_rate": 0.00234375, "reward/invalid_output_rate": 0.00234375, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.02578125, "reward/mean_colbert": 0.97109375, "reward/mean_mrr": 0.9875, "reward/mean_ndcg_at_10": 0.86328125, "reward/mean_r_ndcg": 0.72734375, "reward/mean_r_ndcg_cutoff": 5.68125, "reward/mean_recall": 0.2140625, "reward/mean_recall_at_10": 0.04814453125, "reward/mean_recall_at_100": 0.2140625, "reward/mean_retrieval_combined_bounded": 0.029168701171875, "reward/mean_retrieval_combined_raw": 0.052642011642456056, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 0.99765625, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 0.99765625, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1992.0, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2885029256343842, "rewards/retrieval_reward/mean": 1.02578125, "rewards/retrieval_reward/std": 0.288671875, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8568931579589845, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.29916331768035886, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.478423690795898, "sampling/raw_vllm_sampling_logp_difference/mean": 0.23275801837444304, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 1995, "step_time": 13.737741730781272 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 1.0, "completions/max_length": 35.6, "completions/max_terminated_length": 0.0, "completions/mean_length": 30.98359375, "completions/mean_terminated_length": 0.0, "completions/min_length": 28.4, "completions/min_terminated_length": 0.0, "entropy": 0.3427703082561493, "epoch": 7.905138339920948, "frac_reward_zero_std": 0.0, "grad_norm": 2.0051536560058594, "learning_rate": 2e-05, "loss": 8.23974609375e-05, "num_tokens": 119472954.0, "reward": 1.017608642578125, "reward/completion_length_max": 35.6, "reward/completion_length_mean": 31.0, "reward/completion_length_min": 28.4, "reward/completion_length_p90": 33.6, "reward/completion_length_p95": 34.0, "reward/duplicate_query_rate": 0.0, "reward/duplicate_ranked_plain_query_rate": 0.0, "reward/invalid_output_rate": 0.0, "reward/length_out_of_range_rate": 0.0, "reward/low_length_query_rate": 0.0, "reward/malformed_query_rate": 0.0, "reward/malformed_ranked_plain_query_rate": 0.0, "reward/mean": 1.01875, "reward/mean_colbert": 0.97578125, "reward/mean_mrr": 1.0, "reward/mean_ndcg_at_10": 0.934375, "reward/mean_r_ndcg": 0.78828125, "reward/mean_r_ndcg_cutoff": 6.15, "reward/mean_recall": 0.22890625, "reward/mean_recall_at_10": 0.048876953125, "reward/mean_recall_at_100": 0.22890625, "reward/mean_retrieval_combined_bounded": 0.01760711669921875, "reward/mean_retrieval_combined_raw": 4.0012598037719724e-05, "reward/output_tag_rate": 1.0, "reward/query_valid_rate": 1.0, "reward/r_ndcg_ndcg10_r100_enabled": 1.0, "reward/ranked_plain_query_valid_rate": 1.0, "reward/retrieval_weight_colbert": 0.5, "reward/retrieval_weight_ndcg10": 0.5, "reward/retrieval_weight_r100": 0.0, "reward/retrieval_weight_r_ndcg": 0.5, "reward/retrieval_weight_recall100": 0.0, "reward/retrieval_weight_step": 1998.4, "reward/search_failure_rate": 0.0, "reward/too_long_query_rate": 0.0, "reward/too_short_query_rate": 0.0, "reward_std": 0.2752158224582672, "rewards/retrieval_reward/mean": 1.01875, "rewards/retrieval_reward/std": 0.275390625, "sampling/importance_sampling_ratio/max": 1.0, "sampling/importance_sampling_ratio/mean": 1.0, "sampling/importance_sampling_ratio/min": 1.0, "sampling/raw_vllm_importance_sampling_ratio/max": 2.8642756938934326, "sampling/raw_vllm_importance_sampling_ratio/mean": 0.28745963871479036, "sampling/raw_vllm_importance_sampling_ratio/min": 0.0, "sampling/raw_vllm_sampling_logp_difference/max": 8.899268341064452, "sampling/raw_vllm_sampling_logp_difference/mean": 0.20295145809650422, "sampling/rescore_vllm_sampling_logprobs/enabled": 1.0, "sampling/sampling_logp_difference/max": 0.0, "sampling/sampling_logp_difference/mean": 0.0, "sampling/vllm_logprob_alignment/first_candidate_mismatch_rate": 0.0, "sampling/vllm_logprob_alignment/missing_sampled_token_rate": 0.0, "step": 2000, "step_time": 13.796730328770355 } ], "logging_steps": 5, "max_steps": 3000, "num_input_tokens_seen": 119472954, "num_train_epochs": 12, "save_steps": 50, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": false }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 32, "trial_name": null, "trial_params": null }