{"val-aux/sciknoweval/reward/worst@16/mean":0.2696952380952381,"val-aux/sciknoweval/score/best@8/mean":0.8311666666666666,"val-aux/sciknoweval/score/maj@8/std":0.1528837415995549,"val-aux/sciknoweval/incorrect_format/worst@16/std":0.04613710025134583,"val-aux/sciknoweval/acc/worst@4/mean":0.39438095238095233,"val-aux/sciknoweval/acc/best@8/mean":0.8311666666666666,"perf/max_memory_reserved_gb":137.505859375,"critic/score/min":0,"self_distillation/empty_target_batch":0.23046875,"val-aux/sciknoweval/incorrect_format/best@2/std":0.012277432195181853,"training/global_step":100,"training/rollout_probs_diff_valid":1,"val-aux/sciknoweval/acc/worst@4/std":0.182545872924148,"val-aux/sciknoweval/acc/best@2/mean":0.6968857142857143,"global_seqlen/mean":85677.25,"critic/returns/mean":-0.025867901742458344,"self_distillation/token_reweight_is_rlrt":1,"val-aux/sciknoweval/reward/worst@16/std":0.08420508892562015,"val-aux/sciknoweval/score/maj@16/std":0.11089304329120943,"timing_s/save_checkpoint":6.956284845247865,"rollout_corr/rollout_is_eff_sample_size":0.9949946665841123,"val-aux/sciknoweval/score/worst@4/mean":0.39438095238095233,"response_length_non_aborted/max":8192,"rollout_corr/rollout_is_std":0.07089795172214508,"num_turns/mean":2,"val-aux/sciknoweval/incorrect_format/maj@2/std":0.03180376195882722,"val-aux/sciknoweval/score/best@4/std":0.1424436179078602,"timing_per_token_ms/update_actor":0.0411384974974895,"val-aux/sciknoweval/incorrect_format/std@16":0.03187742018839054,"val-aux/sciknoweval/acc/best@4/mean":0.7765571428571428,"val-aux/sciknoweval/reward/best@2/std":0.20627882758717012,"val-aux/sciknoweval/incorrect_format/worst@8/mean":0.9419428571428571,"val-aux/sciknoweval/reward/maj@16/std":0.11089304329120943,"val-aux/sciknoweval/reward/best@4/mean":0.7765571428571428,"training/rollout_probs_diff_std":0.03298613056540489,"val-core/sciknoweval/acc/mean@16":0.5916666666666667,"prompt_length/clip_ratio":0,"_wandb":{"runtime":8084},"val-aux/sciknoweval/reward/maj@2/mean":0.5911,"self_distillation/success_group_fraction":0.78125,"self_distillation/token_reweight_context_sample_fraction":0.76953125,"val-aux/sciknoweval/score/best@2/mean":0.6968857142857143,"perf/total_num_tokens":685418,"self_distillation/token_reweight_correct_sample_fraction":0.58203125,"response_length_non_aborted/mean":2381.2578125,"rollout_corr/rollout_is_seq_std":0.0032133334316313267,"val-aux/sciknoweval/score/worst@2/std":0.23108433720880842,"val-aux/sciknoweval/reward/worst@2/std":0.23108433720880842,"critic/returns/max":0.875,"timing_s/start_profile":8.443742990493774e-05,"rollout_corr/rollout_is_seq_fraction_low":0,"training/rollout_probs_diff_mean":0.0073742191307246685,"rollout_corr/rollout_is_seq_min":0.991420328617096,"critic/returns/min":-0.875,"val-aux/num_turns/min":2,"val-aux/sciknoweval/score/best@4/mean":0.7765571428571428,"val-aux/sciknoweval/acc/maj@2/std":0.25693213561174044,"perf/throughput":779.2287777281216,"self_distillation/token_reweight_lambda":0.5,"rollout_corr/rollout_ppl":1.0910621960647404,"val-aux/sciknoweval/reward/best@2/mean":0.6968857142857143,"timing_s/agent_loop/generate_sequences/max":69.66160243563354,"val-aux/sciknoweval/acc/maj@8/mean":0.6276761904761905,"rollout_corr/log_ppl_abs_diff":0.006301159230133635,"val-aux/sciknoweval/acc/worst@16/std":0.08420508892562015,"val-aux/sciknoweval/acc/worst@8/mean":0.32224761904761906,"rollout_corr/training_log_ppl":0.0930957942691748,"val-aux/sciknoweval/score/worst@8/std":0.13204132054783738,"val-aux/sciknoweval/reward/best@16/mean":0.8632047619047617,"timing_s/step":109.9513422101736,"val-core/sciknoweval/acc/maj@16/mean":0.6312809523809525,"rollout_corr/rollout_is_min":8.033764970605262e-06,"self_distillation/token_reweight_w_std":1.7234428022899628e+06,"training/rollout_actor_probs_pearson_corr":0.9742870330810547,"val-aux/sciknoweval/incorrect_format/maj@16/mean":0.9972333333333334,"critic/advantages/mean":-0.025867901742458344,"val-aux/num_turns/mean":2,"rollout_corr/log_ppl_diff":0.0062478416239173384,"val-aux/sciknoweval/acc/worst@8/std":0.13204132054783738,"critic/score/max":1,"self_distillation/feedback_available_fraction":0,"response/aborted_ratio":0,"response_length/min":384,"val-aux/sciknoweval/acc/worst@16/mean":0.2696952380952381,"val-aux/sciknoweval/score/std@16":0.2571416687612349,"actor/ppo_kl":0.0011108497318623733,"val-aux/sciknoweval/reward/worst@8/mean":0.32224761904761906,"val-aux/sciknoweval/incorrect_format/worst@16/mean":0.9171,"global_seqlen/minmax_diff":82392,"val-aux/sciknoweval/incorrect_format/maj@16/std":0.005810883427479756,"actor/pg_clipfrac":0.0037484983022295637,"val-aux/sciknoweval/reward/maj@8/mean":0.6276761904761905,"timing_s/agent_loop/generate_sequences/min":4.50063961930573,"rollout_corr/log_ppl_diff_max":0.026261113584041595,"val-aux/sciknoweval/incorrect_format/maj@8/std":0.012077957780667586,"timing_s/agent_loop/slowest/prompt_length":453,"perf/time_per_step":109.9513422101736,"val-aux/sciknoweval/reward/worst@4/mean":0.39438095238095233,"timing_s/agent_loop/tool_calls/max":0,"val-aux/sciknoweval/reward/std@16":0.2571416687612349,"val-aux/sciknoweval/reward/worst@4/std":0.182545872924148,"val-aux/sciknoweval/acc/best@8/std":0.08831510448891618,"rollout_corr/rollout_is_seq_max_deviation":0.025575757026672363,"self_distillation/feedback_used_fraction":0,"val-aux/sciknoweval/incorrect_format/best@4/mean":0.9985428571428571,"val-aux/sciknoweval/acc/maj@4/std":0.20756433319233475,"val-aux/sciknoweval/incorrect_format/best@16/mean":0.9999714285714286,"timing_s/update_actor":28.197066677734256,"response_length_non_aborted/clip_ratio":0.046875,"val-aux/sciknoweval/reward/worst@2/mean":0.4862857142857143,"val-aux/sciknoweval/acc/best@4/std":0.1424436179078602,"rollout_corr/kl":0.006247840840210017,"rollout_corr/chi2_token":0.01678405632264912,"val-aux/sciknoweval/score/maj@2/mean":0.5911,"rollout_corr/rollout_log_ppl":0.08684795264525746,"critic/rewards/mean":0.58203125,"self_distillation/token_reweight_w_clipped_mean":1.0319236246868968,"val-aux/sciknoweval/score/maj@4/std":0.20756433319233475,"critic/advantages/min":-0.875,"val-aux/sciknoweval/incorrect_format/maj@2/mean":0.9870571428571429,"val-aux/sciknoweval/acc/maj@4/mean":0.6155428571428572,"rollout_corr/rollout_is_max":40.18333435058594,"global_seqlen/balanced_min":85577,"val-aux/sciknoweval/reward/worst@8/std":0.13204132054783738,"val-aux/sciknoweval/acc/maj@2/mean":0.5911,"self_distillation/token_reweight_active_token_fraction":0.5703125,"val-aux/sciknoweval/reward/best@4/std":0.1424436179078602,"val-aux/num_turns/max":2,"actor/pg_loss":-0.01704463444184512,"val-aux/sciknoweval/score/maj@2/std":0.25693213561174044,"self_distillation/token_reweight_w_clip_frac":0.056247425003675744,"val-aux/sciknoweval/incorrect_format/worst@4/std":0.048040385516218594,"num_turns/max":2,"rollout_corr/rollout_is_mean":0.9999975562095642,"val-aux/sciknoweval/score/best@8/std":0.08831510448891618,"perf/mfu/actor":0.09517764608800347,"val-aux/sciknoweval/incorrect_format/worst@2/std":0.0401539474232565,"timing_s/gen":76.14493312500417,"val-core/sciknoweval/acc/best@16/std":0.04720106071281998,"val-aux/sciknoweval/acc/worst@2/std":0.23108433720880842,"timing_per_token_ms/adv":0.0024435770715063795,"self_distillation/correct_sample_fraction":0.58203125,"response_length/max":8192,"rollout_corr/rollout_is_seq_mean":0.9998775720596313,"val-aux/sciknoweval/score/best@16/std":0.04720106071281998,"val-aux/sciknoweval/incorrect_format/mean@16":0.986904761904762,"prompt_length/max":625,"val-aux/sciknoweval/incorrect_format/worst@4/mean":0.9634142857142858,"val-aux/sciknoweval/score/maj@8/mean":0.6276761904761905,"val-aux/sciknoweval/incorrect_format/best@8/mean":0.999552380952381,"prompt_length/min":165,"self_distillation/reprompt_sample_fraction":0.76953125,"val-aux/sciknoweval/score/worst@16/mean":0.2696952380952381,"rollout_corr/rollout_is_ratio_fraction_low":0.00278214318677783,"val-aux/sciknoweval/reward/maj@8/std":0.1528837415995549,"global_seqlen/max":133724,"timing_s/agent_loop/tool_calls/mean":0,"_timestamp":1.7834258816012614e+09,"timing_s/reward":0.15107932314276695,"timing_s/old_log_prob":3.69563240557909,"val-aux/sciknoweval/reward/maj@4/std":0.20756433319233475,"val-aux/sciknoweval/score/best@2/std":0.20627882758717012,"response_length_non_aborted/min":384,"rollout_corr/rollout_is_seq_fraction_high":0,"val-aux/sciknoweval/incorrect_format/worst@8/std":0.05198094275109466,"timing_s/agent_loop/slowest/tool_calls":0,"_step":100,"num_turns/min":2,"_runtime":8084.998723123,"val-aux/sciknoweval/score/maj@16/mean":0.6312809523809525,"timing_s/stop_profile":4.6020373702049255e-05,"val-core/sciknoweval/acc/best@16/mean":0.8632047619047617,"rollout_corr/ppl_ratio":1.006276834057644,"rollout_corr/k3_kl":0.006140935789517243,"val-aux/sciknoweval/reward/maj@2/std":0.25693213561174044,"perf/max_memory_allocated_gb":132.60013580322266,"prompt_length/mean":296.15625,"val-aux/sciknoweval/score/maj@4/mean":0.6155428571428572,"rollout_corr/log_ppl_diff_min":-0.0033704042434692383,"perf/mfu/actor_infer":0,"rollout_corr/rollout_is_seq_max":1.0255757570266724,"critic/score/mean":0.58203125,"val-aux/sciknoweval/reward/best@8/std":0.08831510448891618,"actor/grad_norm":0.5481992661952972,"val-aux/sciknoweval/incorrect_format/best@8/std":0.0013896605566054733,"val-aux/sciknoweval/reward/maj@16/mean":0.6312809523809525,"critic/advantages/max":0.875,"timing_s/adv":1.6748717091977596,"val-aux/sciknoweval/reward/mean@16":0.5916666666666667,"val-aux/sciknoweval/score/mean@16":0.5916666666666667,"self_distillation/token_reweight_eps_w":0.5,"val-aux/sciknoweval/score/worst@16/std":0.08420508892562015,"self_distillation/token_reweight_reward_gate_fraction":0.58203125,"val-aux/sciknoweval/score/worst@2/mean":0.4862857142857143,"val-aux/sciknoweval/acc/best@2/std":0.20627882758717012,"training/epoch":1,"val-aux/sciknoweval/incorrect_format/worst@2/mean":0.9777,"critic/rewards/max":1,"timing_s/agent_loop/generate_sequences/mean":22.968712450667226,"critic/rewards/min":0,"actor/kl_loss":0,"global_seqlen/min":51332,"timing_s/agent_loop/tool_calls/min":0,"timing_s/agent_loop/slowest/generate_sequences":69.66160243563354,"self_distillation/success_sample_fraction":0.76953125,"val-aux/sciknoweval/incorrect_format/best@4/std":0.002894135661555999,"val-aux/sciknoweval/incorrect_format/maj@4/mean":0.9926047619047619,"val-aux/sciknoweval/incorrect_format/best@16/std":0.00036774732526300623,"val-aux/sciknoweval/reward/best@8/mean":0.8311666666666666,"val-aux/sciknoweval/incorrect_format/best@2/mean":0.9961809523809524,"global_seqlen/balanced_max":85753,"rollout_corr/rollout_is_ratio_fraction_high":0.00089730677427724,"val-aux/sciknoweval/reward/best@16/std":0.04720106071281998,"val-aux/sciknoweval/acc/maj@8/std":0.1528837415995549,"actor/lr":1e-06,"val-aux/sciknoweval/score/best@16/mean":0.8632047619047617,"val-core/sciknoweval/acc/maj@16/std":0.11089304329120943,"val-aux/sciknoweval/reward/maj@4/mean":0.6155428571428572,"response_length/mean":2381.2578125,"val-aux/sciknoweval/score/worst@4/std":0.182545872924148,"val-aux/sciknoweval/acc/std@16":0.2571416687612349,"rollout_corr/training_ppl":1.0979263153858483,"val-aux/sciknoweval/incorrect_format/maj@4/std":0.021344027228617606,"self_distillation/token_reweight_w_mean":55384.86596345715,"response_length/clip_ratio":0.046875,"val-aux/sciknoweval/score/worst@8/mean":0.32224761904761906,"timing_s/testing":307.67817659303546,"perf/cpu_memory_used_gb":150.47185134887695,"rollout_corr/chi2_seq":32.379441073164344,"val-aux/sciknoweval/incorrect_format/maj@8/mean":0.9957190476190476,"timing_s/agent_loop/slowest/response_length":8192,"training/rollout_probs_diff_max":0.9959022998809814,"val-aux/sciknoweval/acc/worst@2/mean":0.4862857142857143,"actor/pg_clipfrac_lower":7.175036256512612e-05,"timing_per_token_ms/gen":0.12490925739253508,"actor/entropy":0.07674972712993622}