diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,18522 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.26143790849673204, + "eval_steps": 500, + "global_step": 80, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "completions/clipped_ratio": 0.0234375, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 1003.0, + "completions/mean_length": 341.7109375, + "completions/mean_terminated_length": 325.3360290527344, + "completions/min_length": 105.0, + "completions/min_terminated_length": 105.0, + "epoch": 0, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.5, + "global/avg_subtask_length": 111.23376623376623, + "global/chosen_num_routing_steps_ratio": { + "1": 0.25, + "2": 0.21875, + "3": 0.40625, + "4": 0.125, + "5": 0.0 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.5, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.37662337662337664, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 0, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 0, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.4675324675324675, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.2987012987012987, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.23376623376623376, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.625, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.375, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3333333333333333, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.16666666666666666, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.35294117647058826, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.35294117647058826, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.29411764705882354, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.5, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.25, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.25, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.46875, + "global/total_cost": 0.0, + "num_tokens": 256699.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.53125, + "simpleqa_verified/avg_subtask_length": 95.42857142857143, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.47058823529411764, + "2": 0.4117647058823529, + "3": 0.11764705882352941, + "4": 0.0, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.46875, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.5357142857142857, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6428571428571429, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07142857142857142, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.2857142857142857, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1111111111111111, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8888888888888888, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.5, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.53125, + "simpleqa_verified/total_cost": 0.0, + "step": 0, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.46875, + "swe_bench/avg_subtask_length": 120.26530612244898, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0, + "2": 0.0, + "3": 0.7333333333333333, + "4": 0.26666666666666666, + "5": 0.0 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.53125, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.2857142857142857, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3673469387755102, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.42857142857142855, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.20408163265306123, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.8, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5333333333333333, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.2, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.26666666666666666, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3333333333333333, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3333333333333333, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3333333333333333, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.5, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.25, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.25, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.40625, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.424593323841691, + "epoch": 0.0032679738562091504, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.7208398580551147, + "learning_rate": 0.0, + "loss": -0.1107, + "reward": 0.71484375, + "reward_std": 0.5910809636116028, + "rewards/MultiTurnTaskReward/mean": 0.71484375, + "rewards/MultiTurnTaskReward/std": 0.686436653137207, + "sampling/importance_sampling_ratio/max": 1.5569819211959839, + "sampling/importance_sampling_ratio/mean": 0.9997034072875977, + "sampling/importance_sampling_ratio/min": 0.645923376083374, + "sampling/sampling_logp_difference/max": 0.4427492618560791, + "sampling/sampling_logp_difference/mean": 0.014182438142597675, + "step": 1 + }, + { + "completions/clipped_ratio": 0.0078125, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 982.0, + "completions/mean_length": 289.7890625, + "completions/mean_terminated_length": 284.00787353515625, + "completions/min_length": 104.0, + "completions/min_terminated_length": 104.0, + "epoch": 0.0032679738562091504, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.5390625, + "global/avg_subtask_length": 112.16571428571429, + "global/chosen_num_routing_steps_ratio": { + "1": 0.15942028985507245, + "2": 0.2608695652173913, + "3": 0.4782608695652174, + "4": 0.08695652173913043, + "5": 0.014492753623188406 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.4609375, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.29714285714285715, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 0, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 0, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.44571428571428573, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.3314285714285714, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.22285714285714286, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.5797101449275363, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.391304347826087, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.028985507246376812, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3793103448275862, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.22413793103448276, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.39655172413793105, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.5714285714285714, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2857142857142857, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.14285714285714285, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.5078125, + "global/total_cost": 0.0, + "num_tokens": 504640.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.5625, + "simpleqa_verified/avg_subtask_length": 97.3768115942029, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.3055555555555556, + "2": 0.4722222222222222, + "3": 0.2222222222222222, + "4": 0.0, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.4375, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.4057971014492754, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6231884057971014, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07246376811594203, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.30434782608695654, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9722222222222222, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.027777777777777776, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.2, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.68, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.375, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.25, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.375, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5625, + "simpleqa_verified/total_cost": 0.0, + "step": 1, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.515625, + "swe_bench/avg_subtask_length": 121.79245283018868, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0, + "2": 0.030303030303030304, + "3": 0.7575757575757576, + "4": 0.18181818181818182, + "5": 0.030303030303030304 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.484375, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.22641509433962265, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.330188679245283, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.16981132075471697, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.15151515151515152, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.8181818181818182, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.030303030303030304, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5151515151515151, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.30303030303030304, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.18181818181818182, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.28125, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4375, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.28125, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.5714285714285714, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2857142857142857, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.14285714285714285, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.453125, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.40200812742114067, + "epoch": 0.006535947712418301, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.9788647890090942, + "learning_rate": 1.6666666666666665e-07, + "loss": -0.0991, + "reward": 0.8046875, + "reward_std": 0.7374948263168335, + "rewards/MultiTurnTaskReward/mean": 0.8046875, + "rewards/MultiTurnTaskReward/std": 0.7456814646720886, + "sampling/importance_sampling_ratio/max": 1.924230694770813, + "sampling/importance_sampling_ratio/mean": 1.0000910758972168, + "sampling/importance_sampling_ratio/min": 0.6257997155189514, + "sampling/sampling_logp_difference/max": 0.6545262336730957, + "sampling/sampling_logp_difference/mean": 0.014496462419629097, + "step": 2 + }, + { + "completions/clipped_ratio": 0.03125, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 1018.0, + "completions/mean_length": 376.3515625, + "completions/mean_terminated_length": 355.45965576171875, + "completions/min_length": 89.0, + "completions/min_terminated_length": 89.0, + "epoch": 0.006535947712418301, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.5625, + "global/avg_subtask_length": 116.02640264026402, + "global/chosen_num_routing_steps_ratio": { + "1": 0.10185185185185185, + "2": 0.18518518518518517, + "3": 0.5740740740740741, + "4": 0.08333333333333333, + "5": 0.05555555555555555 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.4375, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.2508250825082508, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 0, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 0, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.35973597359735976, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.43234323432343236, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.2079207920792079, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.46296296296296297, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.5092592592592593, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.027777777777777776, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.30927835051546393, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3402061855670103, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.35051546391752575, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2857142857142857, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4155844155844156, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.2987012987012987, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.4, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5333333333333333, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.06666666666666667, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3333333333333333, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.4010416666666667, + "global/total_cost": 0.0, + "num_tokens": 814637.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.5625, + "simpleqa_verified/avg_subtask_length": 97.3768115942029, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.3055555555555556, + "2": 0.4722222222222222, + "3": 0.2222222222222222, + "4": 0.0, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.4375, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.4057971014492754, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6231884057971014, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07246376811594203, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.30434782608695654, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9722222222222222, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.027777777777777776, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.2, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.68, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.375, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.25, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.375, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5625, + "simpleqa_verified/total_cost": 0.0, + "step": 2, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.5625, + "swe_bench/avg_subtask_length": 121.52564102564102, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0, + "2": 0.041666666666666664, + "3": 0.75, + "4": 0.125, + "5": 0.08333333333333333 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.4375, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.20512820512820512, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.28205128205128205, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5384615384615384, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1794871794871795, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.20833333333333334, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7638888888888888, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.027777777777777776, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3472222222222222, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4166666666666667, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.2361111111111111, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2753623188405797, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.43478260869565216, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.2898550724637681, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.4, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5333333333333333, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.06666666666666667, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3333333333333333, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.3203125, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.44883950613439083, + "epoch": 0.00980392156862745, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.846208930015564, + "learning_rate": 3.333333333333333e-07, + "loss": -0.1644, + "reward": 0.71484375, + "reward_std": 0.5736750364303589, + "rewards/MultiTurnTaskReward/mean": 0.71484375, + "rewards/MultiTurnTaskReward/std": 0.6690091490745544, + "sampling/importance_sampling_ratio/max": 1.6813299655914307, + "sampling/importance_sampling_ratio/mean": 1.0000519752502441, + "sampling/importance_sampling_ratio/min": 0.5081575512886047, + "sampling/sampling_logp_difference/max": 0.676963746547699, + "sampling/sampling_logp_difference/mean": 0.015695886686444283, + "step": 3 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.828125, + "aime/avg_subtask_length": 138.1, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.018867924528301886, + "2": 0.1320754716981132, + "3": 0.6792452830188679, + "4": 0.1509433962264151, + "5": 0.018867924528301886 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.171875, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.01875, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 0, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 0, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.225, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01875, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.75625, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.018867924528301886, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9811320754716981, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.28846153846153844, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.019230769230769232, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6923076923076923, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3333333333333333, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.044444444444444446, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6222222222222222, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.5555555555555556, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.4444444444444444, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.765625, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.015625, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 1017.0, + "completions/mean_length": 449.78125, + "completions/mean_terminated_length": 440.66668701171875, + "completions/min_length": 142.0, + "completions/min_terminated_length": 142.0, + "epoch": 0.00980392156862745, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.62890625, + "global/avg_subtask_length": 123.65442764578833, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07453416149068323, + "2": 0.16770186335403728, + "3": 0.6086956521739131, + "4": 0.10559006211180125, + "5": 0.043478260869565216 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.37109375, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.17062634989200864, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 0, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 0, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.31317494600431967, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.2894168466522678, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.39740820734341253, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3167701863354037, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.3416149068322981, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3416149068322981, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.30201342281879195, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.22818791946308725, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4697986577181208, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.30327868852459017, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2786885245901639, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4180327868852459, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.4583333333333333, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3333333333333333, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20833333333333334, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.14285714285714285, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.42857142857142855, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.42857142857142855, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.4921875, + "global/total_cost": 0.0, + "num_tokens": 1059537.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.5625, + "simpleqa_verified/avg_subtask_length": 97.3768115942029, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.3055555555555556, + "2": 0.4722222222222222, + "3": 0.2222222222222222, + "4": 0.0, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.4375, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.4057971014492754, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6231884057971014, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07246376811594203, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.30434782608695654, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9722222222222222, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.027777777777777776, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.2, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.68, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.375, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.25, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.375, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5625, + "simpleqa_verified/total_cost": 0.0, + "step": 3, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.5625, + "swe_bench/avg_subtask_length": 121.52564102564102, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0, + "2": 0.041666666666666664, + "3": 0.75, + "4": 0.125, + "5": 0.08333333333333333 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.4375, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.20512820512820512, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.28205128205128205, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5384615384615384, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1794871794871795, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.20833333333333334, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7638888888888888, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.027777777777777776, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3472222222222222, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4166666666666667, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.2361111111111111, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2753623188405797, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.43478260869565216, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.2898550724637681, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.4, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5333333333333333, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.06666666666666667, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3333333333333333, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.3203125, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.5095600932836533, + "epoch": 0.013071895424836602, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.8018406629562378, + "learning_rate": 5e-07, + "loss": 0.1168, + "reward": 1.08984375, + "reward_std": 0.5611029267311096, + "rewards/MultiTurnTaskReward/mean": 1.08984375, + "rewards/MultiTurnTaskReward/std": 0.596688985824585, + "sampling/importance_sampling_ratio/max": 1.7979037761688232, + "sampling/importance_sampling_ratio/mean": 0.9999969005584717, + "sampling/importance_sampling_ratio/min": 0.6762927174568176, + "sampling/sampling_logp_difference/max": 0.5866214036941528, + "sampling/sampling_logp_difference/mean": 0.015259829349815845, + "step": 4 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.8541666666666666, + "aime/avg_subtask_length": 141.46218487394958, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.012195121951219513, + "2": 0.1951219512195122, + "3": 0.6829268292682927, + "4": 0.0975609756097561, + "5": 0.012195121951219513 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.14583333333333334, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.025210084033613446, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 0, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 0, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.23109243697478993, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.012605042016806723, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7563025210084033, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.024390243902439025, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.975609756097561, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.345679012345679, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.012345679012345678, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6419753086419753, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3076923076923077, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03076923076923077, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6615384615384615, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.5555555555555556, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.4444444444444444, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8125, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0234375, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 836.0, + "completions/mean_length": 393.1171875, + "completions/mean_terminated_length": 377.97601318359375, + "completions/min_length": 94.0, + "completions/min_terminated_length": 94.0, + "epoch": 0.013071895424836602, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.66875, + "global/avg_subtask_length": 125.41004862236629, + "global/chosen_num_routing_steps_ratio": { + "1": 0.056074766355140186, + "2": 0.1822429906542056, + "3": 0.6308411214953271, + "4": 0.08411214953271028, + "5": 0.04672897196261682 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.33125, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.14424635332252836, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 0, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 0, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30307941653160453, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.2787682333873582, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4181523500810373, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.26635514018691586, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.34579439252336447, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3878504672897196, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3564356435643564, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.19801980198019803, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.44554455445544555, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.27607361963190186, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.25766871165644173, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4662576687116564, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.39285714285714285, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.39285714285714285, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.21428571428571427, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.559375, + "global/total_cost": 0.0, + "num_tokens": 1333824.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.5625, + "simpleqa_verified/avg_subtask_length": 97.3768115942029, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.3055555555555556, + "2": 0.4722222222222222, + "3": 0.2222222222222222, + "4": 0.0, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.4375, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.4057971014492754, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6231884057971014, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07246376811594203, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.30434782608695654, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9722222222222222, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.027777777777777776, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.2, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.68, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.375, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.25, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.375, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5625, + "simpleqa_verified/total_cost": 0.0, + "step": 4, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6, + "swe_bench/avg_subtask_length": 119.3258064516129, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0, + "2": 0.0625, + "3": 0.7395833333333334, + "4": 0.10416666666666667, + "5": 0.09375 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.4, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1774193548387097, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2870967741935484, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5290322580645161, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.18387096774193548, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.20833333333333334, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7708333333333334, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.020833333333333332, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.40625, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.375, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.21875, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.24444444444444444, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4222222222222222, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3333333333333333, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3157894736842105, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5789473684210527, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.10526315789473684, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2222222222222222, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5555555555555556, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2222222222222222, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.40625, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.439678555354476, + "epoch": 0.016339869281045753, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.6680312156677246, + "learning_rate": 6.666666666666666e-07, + "loss": 0.0068, + "reward": 1.18359375, + "reward_std": 0.5572637915611267, + "rewards/MultiTurnTaskReward/mean": 1.18359375, + "rewards/MultiTurnTaskReward/std": 0.6143657565116882, + "sampling/importance_sampling_ratio/max": 1.7711583375930786, + "sampling/importance_sampling_ratio/mean": 1.0001912117004395, + "sampling/importance_sampling_ratio/min": 0.40443360805511475, + "sampling/sampling_logp_difference/max": 0.9052677154541016, + "sampling/sampling_logp_difference/mean": 0.014899732545018196, + "step": 5 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.875, + "aime/avg_subtask_length": 132.10591900311528, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.017857142857142856, + "2": 0.22321428571428573, + "3": 0.6428571428571429, + "4": 0.10714285714285714, + "5": 0.008928571428571428 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.125, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.024922118380062305, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 0, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 0, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2336448598130841, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.012461059190031152, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7538940809968847, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.026785714285714284, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9732142857142857, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.38181818181818183, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.00909090909090909, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6090909090909091, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.25882352941176473, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03529411764705882, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.7058823529411765, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.6153846153846154, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.38461538461538464, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8203125, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0078125, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 734.0, + "completions/mean_length": 306.484375, + "completions/mean_terminated_length": 300.83465576171875, + "completions/min_length": 81.0, + "completions/min_terminated_length": 81.0, + "epoch": 0.016339869281045753, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.671875, + "global/avg_subtask_length": 121.41078838174273, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07364341085271318, + "2": 0.2131782945736434, + "3": 0.5891472868217055, + "4": 0.08527131782945736, + "5": 0.03875968992248062 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.328125, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.1507607192254495, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 0, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 0, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30982019363762103, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.24204702627939143, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.44813278008298757, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.27906976744186046, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2868217054263566, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.43410852713178294, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3723849372384937, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.17573221757322174, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.45188284518828453, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2554347826086957, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.23369565217391305, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5108695652173914, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.4375, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.34375, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.21875, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.5364583333333334, + "global/total_cost": 0.0, + "num_tokens": 1549182.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.5208333333333334, + "simpleqa_verified/avg_subtask_length": 91.1195652173913, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.34, + "2": 0.48, + "3": 0.18, + "4": 0.0, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.4791666666666667, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.5, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6521739130434783, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07608695652173914, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.2717391304347826, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.98, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.02, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.24242424242424243, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.15151515151515152, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6060606060606061, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3333333333333333, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2222222222222222, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4444444444444444, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.375, + "simpleqa_verified/total_cost": 0.0, + "step": 5, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6, + "swe_bench/avg_subtask_length": 119.3258064516129, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0, + "2": 0.0625, + "3": 0.7395833333333334, + "4": 0.10416666666666667, + "5": 0.09375 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.4, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1774193548387097, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2870967741935484, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5290322580645161, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.18387096774193548, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.20833333333333334, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7708333333333334, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.020833333333333332, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.40625, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.375, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.21875, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.24444444444444444, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4222222222222222, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3333333333333333, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3157894736842105, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5789473684210527, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.10526315789473684, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2222222222222222, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5555555555555556, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2222222222222222, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.40625, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4402481522411108, + "epoch": 0.0196078431372549, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.8471628427505493, + "learning_rate": 8.333333333333333e-07, + "loss": -0.0354, + "reward": 0.80078125, + "reward_std": 0.3276734948158264, + "rewards/MultiTurnTaskReward/mean": 0.80078125, + "rewards/MultiTurnTaskReward/std": 0.6470400094985962, + "sampling/importance_sampling_ratio/max": 1.579885482788086, + "sampling/importance_sampling_ratio/mean": 0.9997164607048035, + "sampling/importance_sampling_ratio/min": 0.2369312047958374, + "sampling/sampling_logp_difference/max": 1.4399855136871338, + "sampling/sampling_logp_difference/mean": 0.014775452204048634, + "step": 6 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.8958333333333334, + "aime/avg_subtask_length": 133.53319919517102, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.011627906976744186, + "2": 0.20348837209302326, + "3": 0.6802325581395349, + "4": 0.09302325581395349, + "5": 0.011627906976744186 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.10416666666666667, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.022132796780684104, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 1, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 6, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.22535211267605634, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.012072434607645875, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7625754527162978, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.040697674418604654, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9593023255813954, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.34705882352941175, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0058823529411764705, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6470588235294118, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2814814814814815, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.037037037037037035, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6814814814814815, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.4444444444444444, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5555555555555556, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7760416666666666, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.015625, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 916.0, + "completions/mean_length": 393.1171875, + "completions/mean_terminated_length": 383.10321044921875, + "completions/min_length": 80.0, + "completions/min_terminated_length": 80.0, + "epoch": 0.0196078431372549, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.7098214285714286, + "global/avg_subtask_length": 124.29365962180201, + "global/chosen_num_routing_steps_ratio": { + "1": 0.059748427672955975, + "2": 0.20440251572327045, + "3": 0.6194968553459119, + "4": 0.08176100628930817, + "5": 0.03459119496855346 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.29017857142857145, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.12458286985539488, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 1, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 6, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2903225806451613, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19688542825361513, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.5127919911012235, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2389937106918239, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.23270440251572327, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.5283018867924528, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.35451505016722407, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.14046822742474915, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5050167224080268, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2692307692307692, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.19230769230769232, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5384615384615384, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3783783783783784, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2972972972972973, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.32432432432432434, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.45454545454545453, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.36363636363636365, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.5580357142857143, + "global/total_cost": 0.0, + "num_tokens": 1820941.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.5208333333333334, + "simpleqa_verified/avg_subtask_length": 91.1195652173913, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.34, + "2": 0.48, + "3": 0.18, + "4": 0.0, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.4791666666666667, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.5, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6521739130434783, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07608695652173914, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.2717391304347826, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.98, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.02, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.24242424242424243, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.15151515151515152, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6060606060606061, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3333333333333333, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2222222222222222, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4444444444444444, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.375, + "simpleqa_verified/total_cost": 0.0, + "step": 6, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6, + "swe_bench/avg_subtask_length": 119.3258064516129, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0, + "2": 0.0625, + "3": 0.7395833333333334, + "4": 0.10416666666666667, + "5": 0.09375 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.4, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1774193548387097, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2870967741935484, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5290322580645161, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.18387096774193548, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.20833333333333334, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7708333333333334, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.020833333333333332, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.40625, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.375, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.21875, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.24444444444444444, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4222222222222222, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3333333333333333, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3157894736842105, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5789473684210527, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.10526315789473684, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2222222222222222, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5555555555555556, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2222222222222222, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.40625, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4522302746772766, + "epoch": 0.02287581699346405, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.7961458563804626, + "learning_rate": 1e-06, + "loss": 0.0256, + "reward": 0.8515625, + "reward_std": 0.5393282175064087, + "rewards/MultiTurnTaskReward/mean": 0.8515625, + "rewards/MultiTurnTaskReward/std": 0.6561621427536011, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0001375675201416, + "sampling/importance_sampling_ratio/min": 0.5128328204154968, + "sampling/sampling_logp_difference/max": 0.7465817928314209, + "sampling/sampling_logp_difference/mean": 0.014683199115097523, + "step": 7 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.89453125, + "aime/avg_subtask_length": 131.96691729323308, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.008733624454148471, + "2": 0.17903930131004367, + "3": 0.7205240174672489, + "4": 0.08296943231441048, + "5": 0.008733624454148471 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.10546875, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.022556390977443608, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 1, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 9, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.22706766917293233, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01804511278195489, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7548872180451128, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.048034934497816595, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9519650655021834, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3700440528634361, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.00881057268722467, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6211453744493393, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.25806451612903225, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.053763440860215055, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6881720430107527, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.38095238095238093, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6190476190476191, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.80078125, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0234375, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 844.0, + "completions/mean_length": 387.421875, + "completions/mean_terminated_length": 372.1440124511719, + "completions/min_length": 143.0, + "completions/min_terminated_length": 143.0, + "epoch": 0.02287581699346405, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.732421875, + "global/avg_subtask_length": 124.77225866916588, + "global/chosen_num_routing_steps_ratio": { + "1": 0.050666666666666665, + "2": 0.18933333333333333, + "3": 0.6533333333333333, + "4": 0.07733333333333334, + "5": 0.029333333333333333 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.267578125, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.10871602624179943, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 1, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 9, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.28116213683223995, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.17150890346766634, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.5473289597000938, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.21333333333333335, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.19733333333333333, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.5893333333333334, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.36797752808988765, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12078651685393259, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5112359550561798, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.256140350877193, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.17543859649122806, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5684210526315789, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.35, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.275, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.375, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.45454545454545453, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.36363636363636365, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.59765625, + "global/total_cost": 0.0, + "num_tokens": 2084963.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.5208333333333334, + "simpleqa_verified/avg_subtask_length": 91.1195652173913, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.34, + "2": 0.48, + "3": 0.18, + "4": 0.0, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.4791666666666667, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.5, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6521739130434783, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07608695652173914, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.2717391304347826, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.98, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.02, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.24242424242424243, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.15151515151515152, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6060606060606061, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3333333333333333, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2222222222222222, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4444444444444444, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.375, + "simpleqa_verified/total_cost": 0.0, + "step": 7, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6, + "swe_bench/avg_subtask_length": 119.3258064516129, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0, + "2": 0.0625, + "3": 0.7395833333333334, + "4": 0.10416666666666667, + "5": 0.09375 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.4, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1774193548387097, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2870967741935484, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5290322580645161, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.18387096774193548, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.20833333333333334, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7708333333333334, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.020833333333333332, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.40625, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.375, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.21875, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.24444444444444444, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4222222222222222, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3333333333333333, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3157894736842105, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5789473684210527, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.10526315789473684, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2222222222222222, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5555555555555556, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2222222222222222, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.40625, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4418244492262602, + "epoch": 0.026143790849673203, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.7509074211120605, + "learning_rate": 9.99934441832816e-07, + "loss": 0.0281, + "reward": 1.04296875, + "reward_std": 0.5470285415649414, + "rewards/MultiTurnTaskReward/mean": 1.04296875, + "rewards/MultiTurnTaskReward/std": 0.6212362051010132, + "sampling/importance_sampling_ratio/max": 1.762594223022461, + "sampling/importance_sampling_ratio/mean": 1.0001472234725952, + "sampling/importance_sampling_ratio/min": 0.5147930979728699, + "sampling/sampling_logp_difference/max": 0.6639901995658875, + "sampling/sampling_logp_difference/mean": 0.015422801487147808, + "step": 8 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.8819444444444444, + "aime/avg_subtask_length": 135.9145183175034, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.007874015748031496, + "2": 0.17716535433070865, + "3": 0.7283464566929134, + "4": 0.07874015748031496, + "5": 0.007874015748031496 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.11805555555555555, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0203527815468114, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 10, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2198100407055631, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.018995929443690638, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7611940298507462, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.04330708661417323, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9566929133858267, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.34523809523809523, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.007936507936507936, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6468253968253969, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.27053140096618356, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.05314009661835749, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6763285024154589, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.36363636363636365, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.045454545454545456, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5909090909090909, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7430555555555556, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 969.0, + "completions/max_terminated_length": 969.0, + "completions/mean_length": 397.7578125, + "completions/mean_terminated_length": 397.7578125, + "completions/min_length": 65.0, + "completions/min_terminated_length": 65.0, + "epoch": 0.026143790849673203, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.7274305555555556, + "global/avg_subtask_length": 127.76357560568087, + "global/chosen_num_routing_steps_ratio": { + "1": 0.045346062052505964, + "2": 0.18615751789976134, + "3": 0.6634844868735084, + "4": 0.07637231503579953, + "5": 0.028639618138424822 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.2725694444444444, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.10526315789473684, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 10, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2723475355054302, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18796992481203006, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.5396825396825397, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.19809069212410502, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21479713603818615, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.5871121718377088, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.355, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.135, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.51, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2608695652173913, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18944099378881987, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5496894409937888, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3181818181818182, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3409090909090909, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3409090909090909, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.25, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4166666666666667, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3333333333333333, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.578125, + "global/total_cost": 0.0, + "num_tokens": 2355844.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.5208333333333334, + "simpleqa_verified/avg_subtask_length": 91.1195652173913, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.34, + "2": 0.48, + "3": 0.18, + "4": 0.0, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.4791666666666667, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.5, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6521739130434783, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07608695652173914, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.2717391304347826, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.98, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.02, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.24242424242424243, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.15151515151515152, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6060606060606061, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3333333333333333, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2222222222222222, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4444444444444444, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.375, + "simpleqa_verified/total_cost": 0.0, + "step": 8, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.5989583333333334, + "swe_bench/avg_subtask_length": 120.60054347826087, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0, + "2": 0.0782608695652174, + "3": 0.7304347826086957, + "4": 0.10434782608695652, + "5": 0.08695652173913043 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.4010416666666667, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1766304347826087, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2826086956521739, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5543478260869565, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.16304347826086957, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.782608695652174, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.017391304347826087, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.40869565217391307, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.40869565217391307, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.1826086956521739, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2358490566037736, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4528301886792453, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3113207547169811, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2727272727272727, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.6363636363636364, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.09090909090909091, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.3, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.4322916666666667, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.47987600788474083, + "epoch": 0.029411764705882353, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.7636247873306274, + "learning_rate": 9.997377845227574e-07, + "loss": 0.0612, + "reward": 1.07421875, + "reward_std": 0.515068531036377, + "rewards/MultiTurnTaskReward/mean": 1.07421875, + "rewards/MultiTurnTaskReward/std": 0.6182584166526794, + "sampling/importance_sampling_ratio/max": 1.7120273113250732, + "sampling/importance_sampling_ratio/mean": 0.9998866319656372, + "sampling/importance_sampling_ratio/min": 0.6077733635902405, + "sampling/sampling_logp_difference/max": 0.5376782417297363, + "sampling/sampling_logp_difference/mean": 0.015990573912858963, + "step": 9 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.8819444444444444, + "aime/avg_subtask_length": 135.9145183175034, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.007874015748031496, + "2": 0.17716535433070865, + "3": 0.7283464566929134, + "4": 0.07874015748031496, + "5": 0.007874015748031496 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.11805555555555555, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0203527815468114, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 10, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2198100407055631, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.018995929443690638, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7611940298507462, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.04330708661417323, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9566929133858267, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.34523809523809523, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.007936507936507936, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6468253968253969, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.27053140096618356, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.05314009661835749, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6763285024154589, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.36363636363636365, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.045454545454545456, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5909090909090909, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7430555555555556, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.015625, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 795.0, + "completions/mean_length": 348.0, + "completions/mean_terminated_length": 337.2698669433594, + "completions/min_length": 104.0, + "completions/min_terminated_length": 104.0, + "epoch": 0.029411764705882353, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.7421875, + "global/avg_subtask_length": 128.1794095382286, + "global/chosen_num_routing_steps_ratio": { + "1": 0.06315789473684211, + "2": 0.21263157894736842, + "3": 0.6294736842105263, + "4": 0.06947368421052631, + "5": 0.02526315789473684 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.2578125, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.10143830431491295, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 10, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30507191521574567, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.17789553368660105, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.5170325510976533, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2926315789473684, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.18947368421052632, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.5178947368421053, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.33707865168539325, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1303370786516854, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5325842696629214, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.27906976744186046, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.19476744186046513, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5261627906976745, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3333333333333333, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3333333333333333, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3333333333333333, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.25, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4166666666666667, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3333333333333333, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.5203125, + "global/total_cost": 0.0, + "num_tokens": 2580836.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.6625, + "simpleqa_verified/avg_subtask_length": 114.69907407407408, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2641509433962264, + "2": 0.44339622641509435, + "3": 0.2830188679245283, + "4": 0.009433962264150943, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.3375, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.25, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6342592592592593, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.0787037037037037, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.28703703703703703, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9905660377358491, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.009433962264150943, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.20512820512820512, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.11538461538461539, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6794871794871795, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4838709677419355, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.25806451612903225, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.25806451612903225, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.225, + "simpleqa_verified/total_cost": 0.0, + "step": 9, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.5989583333333334, + "swe_bench/avg_subtask_length": 120.60054347826087, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0, + "2": 0.0782608695652174, + "3": 0.7304347826086957, + "4": 0.10434782608695652, + "5": 0.08695652173913043 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.4010416666666667, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1766304347826087, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2826086956521739, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5543478260869565, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.16304347826086957, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.782608695652174, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.017391304347826087, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.40869565217391307, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.40869565217391307, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.1826086956521739, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2358490566037736, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4528301886792453, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3113207547169811, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2727272727272727, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.6363636363636364, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.09090909090909091, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.3, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.4322916666666667, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4269086569547653, + "epoch": 0.032679738562091505, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.7232716083526611, + "learning_rate": 9.994100796397953e-07, + "loss": 0.0869, + "reward": 0.83984375, + "reward_std": 0.3454170227050781, + "rewards/MultiTurnTaskReward/mean": 0.83984375, + "rewards/MultiTurnTaskReward/std": 0.5609805583953857, + "sampling/importance_sampling_ratio/max": 1.7816457748413086, + "sampling/importance_sampling_ratio/mean": 1.0000555515289307, + "sampling/importance_sampling_ratio/min": 0.5432081818580627, + "sampling/sampling_logp_difference/max": 0.6102626323699951, + "sampling/sampling_logp_difference/mean": 0.014050441794097424, + "step": 10 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.890625, + "aime/avg_subtask_length": 136.30565583634177, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.007017543859649123, + "2": 0.16842105263157894, + "3": 0.7368421052631579, + "4": 0.07719298245614035, + "5": 0.010526315789473684 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.109375, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.019253910950661854, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 10, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2190132370637786, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.018050541516245487, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.762936221419976, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03859649122807018, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9614035087719298, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3392226148409894, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.007067137809187279, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6537102473498233, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2851063829787234, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04680851063829787, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6680851063829787, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.32, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.08, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.75625, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.046875, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 905.0, + "completions/mean_length": 389.515625, + "completions/mean_terminated_length": 358.31146240234375, + "completions/min_length": 111.0, + "completions/min_terminated_length": 111.0, + "epoch": 0.032679738562091505, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.7485795454545454, + "global/avg_subtask_length": 127.85936443542934, + "global/chosen_num_routing_steps_ratio": { + "1": 0.06072106261859583, + "2": 0.2030360531309298, + "3": 0.6337760910815939, + "4": 0.07400379506641366, + "5": 0.028462998102466792 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.25142045454545453, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.09330628803245436, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 10, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.29749830966869506, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18593644354293443, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.5165652467883706, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2656546489563567, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.20872865275142316, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.5256166982922201, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3434343434343434, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1292929292929293, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5272727272727272, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.27835051546391754, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.20103092783505155, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.520618556701031, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3333333333333333, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3333333333333333, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3333333333333333, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.26666666666666666, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.3333333333333333, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.4, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.5369318181818182, + "global/total_cost": 0.0, + "num_tokens": 2907142.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.6625, + "simpleqa_verified/avg_subtask_length": 114.69907407407408, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2641509433962264, + "2": 0.44339622641509435, + "3": 0.2830188679245283, + "4": 0.009433962264150943, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.3375, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.25, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6342592592592593, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.0787037037037037, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.28703703703703703, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9905660377358491, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.009433962264150943, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.20512820512820512, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.11538461538461539, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6794871794871795, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4838709677419355, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.25806451612903225, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.25806451612903225, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.225, + "simpleqa_verified/total_cost": 0.0, + "step": 10, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6071428571428571, + "swe_bench/avg_subtask_length": 118.19212962962963, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.014705882352941176, + "2": 0.08823529411764706, + "3": 0.6911764705882353, + "4": 0.11764705882352941, + "5": 0.08823529411764706 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.39285714285714285, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1574074074074074, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2800925925925926, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5625, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1574074074074074, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.17647058823529413, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.8088235294117647, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.014705882352941176, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.43283582089552236, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.39552238805970147, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.17164179104477612, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.21311475409836064, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.48360655737704916, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.30327868852459017, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.32142857142857145, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5714285714285714, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.10714285714285714, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.3333333333333333, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4166666666666667, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.25, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.44642857142857145, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.44998032227158546, + "epoch": 0.03594771241830065, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.7535523772239685, + "learning_rate": 9.989514131188558e-07, + "loss": 0.0859, + "reward": 0.8515625, + "reward_std": 0.5400842428207397, + "rewards/MultiTurnTaskReward/mean": 0.8515625, + "rewards/MultiTurnTaskReward/std": 0.6651012897491455, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0000860691070557, + "sampling/importance_sampling_ratio/min": 0.3721732795238495, + "sampling/sampling_logp_difference/max": 0.9883956909179688, + "sampling/sampling_logp_difference/mean": 0.015432164072990417, + "step": 11 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9036458333333334, + "aime/avg_subtask_length": 136.59014778325124, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.005763688760806916, + "2": 0.16138328530259366, + "3": 0.7463976945244957, + "4": 0.07492795389048991, + "5": 0.011527377521613832 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.09635416666666667, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.016748768472906402, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 10, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.22167487684729065, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.020689655172413793, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7576354679802956, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.040345821325648415, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9596541786743515, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3333333333333333, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.005797101449275362, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6608695652173913, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3044982698961938, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.058823529411764705, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6366782006920415, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.26666666666666666, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.06666666666666667, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6666666666666666, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7369791666666666, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 875.0, + "completions/max_terminated_length": 875.0, + "completions/mean_length": 380.109375, + "completions/mean_terminated_length": 380.109375, + "completions/min_length": 69.0, + "completions/min_terminated_length": 69.0, + "epoch": 0.03594771241830065, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.7669270833333334, + "global/avg_subtask_length": 128.96752856283825, + "global/chosen_num_routing_steps_ratio": { + "1": 0.05432937181663837, + "2": 0.19524617996604415, + "3": 0.6502546689303905, + "4": 0.0730050933786078, + "5": 0.027164685908319185 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.23307291666666666, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.0835838845460012, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 10, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2904389657245941, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.16897173782321107, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.5405892964521948, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2427843803056027, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.1867572156196944, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.5704584040747029, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3393177737881508, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.11490125673249552, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5457809694793537, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2918552036199095, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.19004524886877827, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5180995475113123, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3050847457627119, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3050847457627119, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3898305084745763, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.25, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.3125, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.4375, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.5455729166666666, + "global/total_cost": 0.0, + "num_tokens": 3191828.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.6625, + "simpleqa_verified/avg_subtask_length": 114.69907407407408, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2641509433962264, + "2": 0.44339622641509435, + "3": 0.2830188679245283, + "4": 0.009433962264150943, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.3375, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.25, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6342592592592593, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.0787037037037037, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.28703703703703703, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9905660377358491, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.009433962264150943, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.20512820512820512, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.11538461538461539, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6794871794871795, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4838709677419355, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.25806451612903225, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.25806451612903225, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.225, + "simpleqa_verified/total_cost": 0.0, + "step": 11, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6071428571428571, + "swe_bench/avg_subtask_length": 118.19212962962963, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.014705882352941176, + "2": 0.08823529411764706, + "3": 0.6911764705882353, + "4": 0.11764705882352941, + "5": 0.08823529411764706 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.39285714285714285, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1574074074074074, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2800925925925926, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5625, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1574074074074074, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.17647058823529413, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.8088235294117647, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.014705882352941176, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.43283582089552236, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.39552238805970147, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.17164179104477612, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.21311475409836064, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.48360655737704916, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.30327868852459017, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.32142857142857145, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5714285714285714, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.10714285714285714, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.3333333333333333, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4166666666666667, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.25, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.44642857142857145, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4815131947398186, + "epoch": 0.0392156862745098, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.6698487997055054, + "learning_rate": 9.983619052372847e-07, + "loss": -0.0637, + "reward": 0.77734375, + "reward_std": 0.3500558137893677, + "rewards/MultiTurnTaskReward/mean": 0.77734375, + "rewards/MultiTurnTaskReward/std": 0.5279905200004578, + "sampling/importance_sampling_ratio/max": 1.6307594776153564, + "sampling/importance_sampling_ratio/mean": 1.0001192092895508, + "sampling/importance_sampling_ratio/min": 0.5869132876396179, + "sampling/sampling_logp_difference/max": 0.5328781604766846, + "sampling/sampling_logp_difference/mean": 0.015364541672170162, + "step": 12 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.90625, + "aime/avg_subtask_length": 137.12295081967213, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.007957559681697613, + "2": 0.16445623342175067, + "3": 0.7453580901856764, + "4": 0.07161803713527852, + "5": 0.010610079575596816 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.09375, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.015482695810564663, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 10, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.22131147540983606, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.020947176684881604, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7577413479052824, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.042440318302387266, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9575596816976127, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.32620320855614976, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.008021390374331552, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6657754010695187, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3108974358974359, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.057692307692307696, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6314102564102564, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.25806451612903225, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.06451612903225806, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6774193548387096, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7427884615384616, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0078125, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 908.0, + "completions/mean_length": 357.109375, + "completions/mean_terminated_length": 351.8582763671875, + "completions/min_length": 115.0, + "completions/min_terminated_length": 115.0, + "epoch": 0.0392156862745098, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.7704326923076923, + "global/avg_subtask_length": 129.13270925110132, + "global/chosen_num_routing_steps_ratio": { + "1": 0.0514820592823713, + "2": 0.19032761310452417, + "3": 0.6599063962558502, + "4": 0.07020280811232449, + "5": 0.028081123244929798 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.22956730769230768, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.08204845814977973, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 10, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2868942731277533, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.17731277533039647, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.5357929515418502, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.23088923556942278, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.1981279251170047, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.5709828393135725, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.34210526315789475, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12006578947368421, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.537828947368421, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.29218106995884774, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.19547325102880658, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5123456790123457, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30158730158730157, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.31746031746031744, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.38095238095238093, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2222222222222222, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.3888888888888889, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3888888888888889, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.5600961538461539, + "global/total_cost": 0.0, + "num_tokens": 3470114.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.6625, + "simpleqa_verified/avg_subtask_length": 114.69907407407408, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2641509433962264, + "2": 0.44339622641509435, + "3": 0.2830188679245283, + "4": 0.009433962264150943, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.3375, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.25, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6342592592592593, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.0787037037037037, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.28703703703703703, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9905660377358491, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.009433962264150943, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.20512820512820512, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.11538461538461539, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6794871794871795, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4838709677419355, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.25806451612903225, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.25806451612903225, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.225, + "simpleqa_verified/total_cost": 0.0, + "step": 12, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6171875, + "swe_bench/avg_subtask_length": 117.86653386454184, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.012658227848101266, + "2": 0.08227848101265822, + "3": 0.7088607594936709, + "4": 0.10759493670886076, + "5": 0.08860759493670886 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.3828125, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1553784860557769, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.28087649402390436, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5617529880478087, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.15737051792828685, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.17088607594936708, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.8037974683544303, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.02531645569620253, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.44871794871794873, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.391025641025641, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.16025641025641027, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2097902097902098, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4825174825174825, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3076923076923077, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3225806451612903, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5806451612903226, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0967741935483871, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2857142857142857, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.21428571428571427, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.47265625, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4466289784759283, + "epoch": 0.042483660130718956, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.7955829501152039, + "learning_rate": 9.97641710583307e-07, + "loss": 0.0168, + "reward": 1.1015625, + "reward_std": 0.6097025871276855, + "rewards/MultiTurnTaskReward/mean": 1.1015625, + "rewards/MultiTurnTaskReward/std": 0.628582239151001, + "sampling/importance_sampling_ratio/max": 1.458682894706726, + "sampling/importance_sampling_ratio/mean": 1.000030279159546, + "sampling/importance_sampling_ratio/min": 0.41387939453125, + "sampling/sampling_logp_difference/max": 0.8821806907653809, + "sampling/sampling_logp_difference/mean": 0.014805847778916359, + "step": 13 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.90625, + "aime/avg_subtask_length": 137.12295081967213, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.007957559681697613, + "2": 0.16445623342175067, + "3": 0.7453580901856764, + "4": 0.07161803713527852, + "5": 0.010610079575596816 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.09375, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.015482695810564663, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 10, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.22131147540983606, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.020947176684881604, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7577413479052824, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.042440318302387266, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9575596816976127, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.32620320855614976, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.008021390374331552, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6657754010695187, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3108974358974359, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.057692307692307696, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6314102564102564, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.25806451612903225, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.06451612903225806, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6774193548387096, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7427884615384616, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 840.0, + "completions/max_terminated_length": 840.0, + "completions/mean_length": 313.484375, + "completions/mean_terminated_length": 313.484375, + "completions/min_length": 116.0, + "completions/min_terminated_length": 116.0, + "epoch": 0.042483660130718956, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.7745535714285714, + "global/avg_subtask_length": 129.2686796315251, + "global/chosen_num_routing_steps_ratio": { + "1": 0.05475504322766571, + "2": 0.1988472622478386, + "3": 0.6498559077809798, + "4": 0.069164265129683, + "5": 0.027377521613832854 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.22544642857142858, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.07830092118730808, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 10, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.29733879222108495, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18730808597748208, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.515353121801433, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.260806916426513, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21181556195965417, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.5273775216138329, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3445121951219512, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12347560975609756, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5320121951219512, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.28957528957528955, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21042471042471042, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.29850746268656714, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.31343283582089554, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3880597014925373, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.21052631578947367, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.42105263157894735, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3684210526315789, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.5457589285714286, + "global/total_cost": 0.0, + "num_tokens": 3733664.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.7083333333333334, + "simpleqa_verified/avg_subtask_length": 122.44483985765125, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2426470588235294, + "2": 0.45588235294117646, + "3": 0.29411764705882354, + "4": 0.007352941176470588, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.2916666666666667, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.199288256227758, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6227758007117438, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.08185053380782918, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.29537366548042704, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9926470588235294, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007352941176470588, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.18446601941747573, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.10679611650485436, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7087378640776699, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4878048780487805, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2926829268292683, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.21951219512195122, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.1875, + "simpleqa_verified/total_cost": 0.0, + "step": 13, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6284722222222222, + "swe_bench/avg_subtask_length": 117.60521739130435, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.011049723756906077, + "2": 0.07734806629834254, + "3": 0.7182320441988951, + "4": 0.11049723756906077, + "5": 0.08287292817679558 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.3715277777777778, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1391304347826087, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.28347826086956524, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5565217391304348, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.16, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.16574585635359115, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.8121546961325967, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.022099447513812154, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4748603351955307, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3743016759776536, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.15083798882681565, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.47878787878787876, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3212121212121212, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3142857142857143, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5428571428571428, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.14285714285714285, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.26666666666666666, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5333333333333333, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.42960745096206665, + "epoch": 0.0457516339869281, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.8281583189964294, + "learning_rate": 9.967910180154888e-07, + "loss": 0.0127, + "reward": 1.01171875, + "reward_std": 0.4565136432647705, + "rewards/MultiTurnTaskReward/mean": 1.01171875, + "rewards/MultiTurnTaskReward/std": 0.5934844613075256, + "sampling/importance_sampling_ratio/max": 1.7732880115509033, + "sampling/importance_sampling_ratio/mean": 1.0001161098480225, + "sampling/importance_sampling_ratio/min": 0.5603508353233337, + "sampling/sampling_logp_difference/max": 0.5791921615600586, + "sampling/sampling_logp_difference/mean": 0.014878196641802788, + "step": 14 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9084821428571429, + "aime/avg_subtask_length": 137.2053646269908, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.007371007371007371, + "2": 0.15724815724815724, + "3": 0.7444717444717445, + "4": 0.07862407862407862, + "5": 0.012285012285012284 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.09151785714285714, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.014249790444258172, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 10, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.22632020117351215, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.019279128248113998, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7544006705783739, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.04668304668304668, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9533169533169533, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.33663366336633666, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.007425742574257425, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.655940594059406, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3058823529411765, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.052941176470588235, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6411764705882353, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2972972972972973, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.05405405405405406, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6486486486486487, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7566964285714286, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0234375, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 877.0, + "completions/mean_length": 384.0703125, + "completions/mean_terminated_length": 368.7120056152344, + "completions/min_length": 87.0, + "completions/min_terminated_length": 87.0, + "epoch": 0.0457516339869281, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.7854166666666667, + "global/avg_subtask_length": 128.13154172560112, + "global/chosen_num_routing_steps_ratio": { + "1": 0.05702917771883289, + "2": 0.1962864721485411, + "3": 0.649867374005305, + "4": 0.07029177718832891, + "5": 0.026525198938992044 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.21458333333333332, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.07260726072607261, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 10, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30787364450730786, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.17491749174917492, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.5172088637435172, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.28249336870026526, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.19496021220159152, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.5225464190981433, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3488045007032349, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.11392405063291139, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5372714486638537, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.29307282415630553, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2024866785079929, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5044404973357016, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3150684931506849, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2876712328767123, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3972602739726027, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.4, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.571875, + "global/total_cost": 0.0, + "num_tokens": 4027753.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.7410714285714286, + "simpleqa_verified/avg_subtask_length": 114.61189801699717, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2289156626506024, + "2": 0.42168674698795183, + "3": 0.3433734939759036, + "4": 0.006024096385542169, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.25892857142857145, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.16147308781869688, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.623229461756374, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07932011331444759, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.29745042492917845, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9879518072289156, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012048192771084338, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.2109375, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0859375, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.703125, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4827586206896552, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.29310344827586204, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.22413793103448276, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.29464285714285715, + "simpleqa_verified/total_cost": 0.0, + "step": 14, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6284722222222222, + "swe_bench/avg_subtask_length": 117.60521739130435, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.011049723756906077, + "2": 0.07734806629834254, + "3": 0.7182320441988951, + "4": 0.11049723756906077, + "5": 0.08287292817679558 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.3715277777777778, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1391304347826087, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.28347826086956524, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5565217391304348, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.16, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.16574585635359115, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.8121546961325967, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.022099447513812154, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4748603351955307, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3743016759776536, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.15083798882681565, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.47878787878787876, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3212121212121212, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3142857142857143, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5428571428571428, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.14285714285714285, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.26666666666666666, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5333333333333333, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.43900963105261326, + "epoch": 0.049019607843137254, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.724708080291748, + "learning_rate": 9.958100506132126e-07, + "loss": 0.068, + "reward": 1.2734375, + "reward_std": 0.38008320331573486, + "rewards/MultiTurnTaskReward/mean": 1.2734375, + "rewards/MultiTurnTaskReward/std": 0.4854552745819092, + "sampling/importance_sampling_ratio/max": 1.954588770866394, + "sampling/importance_sampling_ratio/mean": 0.9999356269836426, + "sampling/importance_sampling_ratio/min": 0.5363368988037109, + "sampling/sampling_logp_difference/max": 0.6701798439025879, + "sampling/sampling_logp_difference/mean": 0.014698440209031105, + "step": 15 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9084821428571429, + "aime/avg_subtask_length": 137.2053646269908, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.007371007371007371, + "2": 0.15724815724815724, + "3": 0.7444717444717445, + "4": 0.07862407862407862, + "5": 0.012285012285012284 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.09151785714285714, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.014249790444258172, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 10, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.22632020117351215, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.019279128248113998, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7544006705783739, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.04668304668304668, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9533169533169533, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.33663366336633666, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.007425742574257425, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.655940594059406, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3058823529411765, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.052941176470588235, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6411764705882353, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2972972972972973, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.05405405405405406, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6486486486486487, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7566964285714286, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.046875, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 964.0, + "completions/mean_length": 375.2265625, + "completions/mean_terminated_length": 343.31964111328125, + "completions/min_length": 52.0, + "completions/min_terminated_length": 52.0, + "epoch": 0.049019607843137254, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.763671875, + "global/avg_subtask_length": 127.45895691609978, + "global/chosen_num_routing_steps_ratio": { + "1": 0.054987212276214836, + "2": 0.18925831202046037, + "3": 0.6624040920716112, + "4": 0.06777493606138107, + "5": 0.02557544757033248 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.236328125, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.08526077097505669, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 10, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3074829931972789, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1854875283446712, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.5070294784580499, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2851662404092072, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21099744245524296, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.5038363171355499, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.34506089309878213, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12313937753721245, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5317997293640054, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2927241962774958, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2098138747884941, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.49746192893401014, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3150684931506849, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2876712328767123, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3972602739726027, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.4, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.560546875, + "global/total_cost": 0.0, + "num_tokens": 4548230.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.7410714285714286, + "simpleqa_verified/avg_subtask_length": 114.61189801699717, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2289156626506024, + "2": 0.42168674698795183, + "3": 0.3433734939759036, + "4": 0.006024096385542169, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.25892857142857145, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.16147308781869688, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.623229461756374, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07932011331444759, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.29745042492917845, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9879518072289156, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012048192771084338, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.2109375, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0859375, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.703125, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4827586206896552, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.29310344827586204, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.22413793103448276, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.29464285714285715, + "simpleqa_verified/total_cost": 0.0, + "step": 15, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.59375, + "swe_bench/avg_subtask_length": 116.69650986342944, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.009569377990430622, + "2": 0.06698564593301436, + "3": 0.7559808612440191, + "4": 0.09569377990430622, + "5": 0.07177033492822966 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.40625, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.17298937784522003, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2852807283763278, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.543247344461305, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17147192716236723, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.19138755980861244, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7894736842105263, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.019138755980861243, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4444444444444444, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3719806763285024, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.18357487922705315, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.21243523316062177, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.46113989637305697, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.32642487046632124, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3142857142857143, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5428571428571428, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.14285714285714285, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.26666666666666666, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5333333333333333, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.48011363636363635, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.41933287493884563, + "epoch": 0.05228758169934641, + "frac_reward_zero_std": 0.25, + "grad_norm": 0.5959435701370239, + "learning_rate": 9.946990656181779e-07, + "loss": -0.0198, + "reward": 0.76171875, + "reward_std": 0.26621413230895996, + "rewards/MultiTurnTaskReward/mean": 0.76171875, + "rewards/MultiTurnTaskReward/std": 0.669836163520813, + "sampling/importance_sampling_ratio/max": 1.673732876777649, + "sampling/importance_sampling_ratio/mean": 1.0001013278961182, + "sampling/importance_sampling_ratio/min": 0.5696284770965576, + "sampling/sampling_logp_difference/max": 0.5627708435058594, + "sampling/sampling_logp_difference/mean": 0.01477130688726902, + "step": 16 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9084821428571429, + "aime/avg_subtask_length": 137.2053646269908, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.007371007371007371, + "2": 0.15724815724815724, + "3": 0.7444717444717445, + "4": 0.07862407862407862, + "5": 0.012285012285012284 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.09151785714285714, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.014249790444258172, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 10, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.22632020117351215, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.019279128248113998, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7544006705783739, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.04668304668304668, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9533169533169533, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.33663366336633666, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.007425742574257425, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.655940594059406, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3058823529411765, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.052941176470588235, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6411764705882353, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2972972972972973, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.05405405405405406, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6486486486486487, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7566964285714286, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0078125, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 696.0, + "completions/mean_length": 318.28125, + "completions/mean_terminated_length": 312.7243957519531, + "completions/min_length": 94.0, + "completions/min_terminated_length": 94.0, + "epoch": 0.05228758169934641, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.7748161764705882, + "global/avg_subtask_length": 125.16673866090713, + "global/chosen_num_routing_steps_ratio": { + "1": 0.08185053380782918, + "2": 0.20166073546856464, + "3": 0.6287069988137604, + "4": 0.06405693950177936, + "5": 0.02372479240806643 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.22518382352941177, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.08250539956803456, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 10, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3257019438444924, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.17926565874730022, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.49503239740820737, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3368920521945433, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.19572953736654805, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.46737841043890865, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3385012919896641, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.11886304909560723, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5426356589147286, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2980132450331126, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21357615894039736, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.48841059602649006, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.32432432432432434, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.28378378378378377, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3918918918918919, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.4, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.5661764705882353, + "global/total_cost": 0.0, + "num_tokens": 4804042.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.7881944444444444, + "simpleqa_verified/avg_subtask_length": 106.20302375809935, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.28193832599118945, + "2": 0.4052863436123348, + "3": 0.3039647577092511, + "4": 0.00881057268722467, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.21180555555555555, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.12958963282937366, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6393088552915767, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.0734341252699784, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.28725701943844495, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9911894273127754, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.00881057268722467, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.2085889570552147, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0736196319018405, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7177914110429447, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.49295774647887325, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.30985915492957744, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.19718309859154928, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.375, + "simpleqa_verified/total_cost": 0.0, + "step": 16, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.59375, + "swe_bench/avg_subtask_length": 116.69650986342944, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.009569377990430622, + "2": 0.06698564593301436, + "3": 0.7559808612440191, + "4": 0.09569377990430622, + "5": 0.07177033492822966 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.40625, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.17298937784522003, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2852807283763278, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.543247344461305, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17147192716236723, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.19138755980861244, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7894736842105263, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.019138755980861243, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4444444444444444, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3719806763285024, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.18357487922705315, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.21243523316062177, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.46113989637305697, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.32642487046632124, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3142857142857143, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5428571428571428, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.14285714285714285, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.26666666666666666, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5333333333333333, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.48011363636363635, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.42122949473559856, + "epoch": 0.05555555555555555, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.7645728588104248, + "learning_rate": 9.934583543669453e-07, + "loss": 0.1093, + "reward": 1.0078125, + "reward_std": 0.47068068385124207, + "rewards/MultiTurnTaskReward/mean": 1.0078125, + "rewards/MultiTurnTaskReward/std": 0.5469470620155334, + "sampling/importance_sampling_ratio/max": 1.718208909034729, + "sampling/importance_sampling_ratio/mean": 0.9999223351478577, + "sampling/importance_sampling_ratio/min": 0.394557386636734, + "sampling/sampling_logp_difference/max": 0.9299907684326172, + "sampling/sampling_logp_difference/mean": 0.015744857490062714, + "step": 17 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9145833333333333, + "aime/avg_subtask_length": 135.5163043478261, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.00683371298405467, + "2": 0.15034168564920272, + "3": 0.7562642369020501, + "4": 0.07517084282460136, + "5": 0.011389521640091117 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.08541666666666667, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.013198757763975156, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 10, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.23214285714285715, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.02096273291925466, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7468944099378882, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.05239179954441914, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9476082004555809, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3532110091743119, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.006880733944954129, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6399082568807339, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.05945945945945946, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6405405405405405, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2894736842105263, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.05263157894736842, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6578947368421053, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 1.0, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7708333333333334, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 751.0, + "completions/max_terminated_length": 751.0, + "completions/mean_length": 365.046875, + "completions/mean_terminated_length": 365.046875, + "completions/min_length": 100.0, + "completions/min_terminated_length": 100.0, + "epoch": 0.05555555555555555, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.7838541666666666, + "global/avg_subtask_length": 124.26294660778804, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07641196013289037, + "2": 0.19490586932447398, + "3": 0.6445182724252492, + "4": 0.06201550387596899, + "5": 0.0221483942414175 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.21614583333333334, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.07747892412685668, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 10, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.32476916900843034, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1834604576475311, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.49177037334403856, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3333333333333333, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.19933554817275748, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4673311184939092, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3501199040767386, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12949640287769784, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5203836930455635, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2857142857142857, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2127659574468085, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5015197568389058, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3157894736842105, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.27631578947368424, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.40789473684210525, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.4, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.5642361111111112, + "global/total_cost": 0.0, + "num_tokens": 5110224.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.7881944444444444, + "simpleqa_verified/avg_subtask_length": 106.20302375809935, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.28193832599118945, + "2": 0.4052863436123348, + "3": 0.3039647577092511, + "4": 0.00881057268722467, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.21180555555555555, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.12958963282937366, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6393088552915767, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.0734341252699784, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.28725701943844495, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9911894273127754, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.00881057268722467, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.2085889570552147, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0736196319018405, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7177914110429447, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.49295774647887325, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.30985915492957744, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.19718309859154928, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.375, + "simpleqa_verified/total_cost": 0.0, + "step": 17, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6171875, + "swe_bench/avg_subtask_length": 115.97567567567567, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.008438818565400843, + "2": 0.0759493670886076, + "3": 0.7637130801687764, + "4": 0.08860759493670886, + "5": 0.06329113924050633 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.3828125, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.15675675675675677, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2891891891891892, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5351351351351351, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17567567567567569, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22362869198312235, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.759493670886076, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.016877637130801686, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4425531914893617, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.39574468085106385, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.16170212765957448, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.1935483870967742, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4423963133640553, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3640552995391705, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3055555555555556, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5277777777777778, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.16666666666666666, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.26666666666666666, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5333333333333333, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.4479166666666667, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4182112533599138, + "epoch": 0.058823529411764705, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.6718185544013977, + "learning_rate": 9.92088242214537e-07, + "loss": 0.0139, + "reward": 0.97265625, + "reward_std": 0.3152496814727783, + "rewards/MultiTurnTaskReward/mean": 0.97265625, + "rewards/MultiTurnTaskReward/std": 0.5516924262046814, + "sampling/importance_sampling_ratio/max": 1.6047571897506714, + "sampling/importance_sampling_ratio/mean": 0.9998775720596313, + "sampling/importance_sampling_ratio/min": 0.3373199701309204, + "sampling/sampling_logp_difference/max": 1.0867233276367188, + "sampling/sampling_logp_difference/mean": 0.015345807187259197, + "step": 18 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9140625, + "aime/avg_subtask_length": 136.4517766497462, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.00641025641025641, + "2": 0.14102564102564102, + "3": 0.7670940170940171, + "4": 0.07051282051282051, + "5": 0.014957264957264958 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.0859375, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.012327773749093546, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 10, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.23060188542422044, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.02102973168963017, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7483683828861494, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.049145299145299144, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9508547008547008, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.34838709677419355, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.008602150537634409, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6430107526881721, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3032581453634085, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.05764411027568922, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6390977443609023, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.275, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.05, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.675, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.14285714285714285, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8571428571428571, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.775390625, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0078125, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 833.0, + "completions/mean_length": 329.4296875, + "completions/mean_terminated_length": 323.96063232421875, + "completions/min_length": 106.0, + "completions/min_terminated_length": 106.0, + "epoch": 0.058823529411764705, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.7911184210526315, + "global/avg_subtask_length": 124.65028355387524, + "global/chosen_num_routing_steps_ratio": { + "1": 0.08108108108108109, + "2": 0.19230769230769232, + "3": 0.6455301455301455, + "4": 0.058212058212058215, + "5": 0.02286902286902287 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.20888157894736842, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.07334593572778828, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 10, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3285444234404537, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.17504725897920606, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.49640831758034026, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.34407484407484407, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.18711018711018712, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4688149688149688, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3427601809954751, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12330316742081449, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5339366515837104, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2947067238912732, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.20743919885550788, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4978540772532189, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3076923076923077, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2692307692307692, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.4230769230769231, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.22727272727272727, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.36363636363636365, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.4090909090909091, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.5797697368421053, + "global/total_cost": 0.0, + "num_tokens": 5376935.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.803125, + "simpleqa_verified/avg_subtask_length": 105.91444866920152, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2840466926070039, + "2": 0.39299610894941633, + "3": 0.3151750972762646, + "4": 0.007782101167315175, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.196875, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.11596958174904944, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6406844106463878, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07224334600760456, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.2870722433460076, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9922178988326849, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007782101167315175, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.20108695652173914, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.06521739130434782, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7336956521739131, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5180722891566265, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3132530120481928, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.1686746987951807, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.425, + "simpleqa_verified/total_cost": 0.0, + "step": 18, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6171875, + "swe_bench/avg_subtask_length": 115.97567567567567, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.008438818565400843, + "2": 0.0759493670886076, + "3": 0.7637130801687764, + "4": 0.08860759493670886, + "5": 0.06329113924050633 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.3828125, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.15675675675675677, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2891891891891892, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5351351351351351, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17567567567567569, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22362869198312235, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.759493670886076, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.016877637130801686, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4425531914893617, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.39574468085106385, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.16170212765957448, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.1935483870967742, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4423963133640553, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3640552995391705, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3055555555555556, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5277777777777778, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.16666666666666666, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.26666666666666666, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5333333333333333, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.4479166666666667, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3994713630527258, + "epoch": 0.06209150326797386, + "frac_reward_zero_std": 0.0, + "grad_norm": 1.2381093502044678, + "learning_rate": 9.905890884491194e-07, + "loss": 0.0093, + "reward": 1.18359375, + "reward_std": 0.49366945028305054, + "rewards/MultiTurnTaskReward/mean": 1.18359375, + "rewards/MultiTurnTaskReward/std": 0.5319370627403259, + "sampling/importance_sampling_ratio/max": 1.6296584606170654, + "sampling/importance_sampling_ratio/mean": 1.0002162456512451, + "sampling/importance_sampling_ratio/min": 0.39963987469673157, + "sampling/sampling_logp_difference/max": 0.9171915054321289, + "sampling/sampling_logp_difference/mean": 0.014847876504063606, + "step": 19 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9191176470588235, + "aime/avg_subtask_length": 136.91032608695653, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.006, + "2": 0.138, + "3": 0.776, + "4": 0.066, + "5": 0.014 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.08088235294117647, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.01154891304347826, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2296195652173913, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.021739130434782608, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.748641304347826, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.046, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.954, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3440643863179074, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.014084507042253521, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.641851106639839, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.308411214953271, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.053738317757009345, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6378504672897196, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.275, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.05, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.675, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.14285714285714285, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8571428571428571, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7683823529411765, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 1018.0, + "completions/max_terminated_length": 1018.0, + "completions/mean_length": 330.4609375, + "completions/mean_terminated_length": 330.4609375, + "completions/min_length": 80.0, + "completions/min_terminated_length": 80.0, + "epoch": 0.06209150326797386, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8015625, + "global/avg_subtask_length": 124.92183098591549, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07602339181286549, + "2": 0.1861598440545809, + "3": 0.6549707602339181, + "4": 0.0594541910331384, + "5": 0.023391812865497075 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.1984375, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.06830985915492958, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3232394366197183, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18274647887323944, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.49401408450704226, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.32943469785575047, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.19980506822612085, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.47076023391812866, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.34177215189873417, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1308016877637131, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5274261603375527, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.29590488771466317, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.20871862615587847, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4953764861294584, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3176470588235294, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.25882352941176473, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.4235294117647059, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.20833333333333334, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4166666666666667, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.375, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.57421875, + "global/total_cost": 0.0, + "num_tokens": 5684610.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.803125, + "simpleqa_verified/avg_subtask_length": 105.91444866920152, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2840466926070039, + "2": 0.39299610894941633, + "3": 0.3151750972762646, + "4": 0.007782101167315175, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.196875, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.11596958174904944, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6406844106463878, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.07224334600760456, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.2870722433460076, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9922178988326849, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007782101167315175, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.20108695652173914, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.06521739130434782, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7336956521739131, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5180722891566265, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3132530120481928, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.1686746987951807, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.425, + "simpleqa_verified/total_cost": 0.0, + "step": 19, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6466346153846154, + "swe_bench/avg_subtask_length": 115.8372921615202, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.007434944237918215, + "2": 0.07806691449814127, + "3": 0.7546468401486989, + "4": 0.09665427509293681, + "5": 0.06319702602230483 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.35336538461538464, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1377672209026128, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.28859857482185275, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5332541567695962, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17814726840855108, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22304832713754646, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7620817843866171, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.01486988847583643, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4344569288389513, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.39325842696629215, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.17228464419475656, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.1991869918699187, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.44308943089430897, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.35772357723577236, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.32558139534883723, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46511627906976744, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20930232558139536, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.23529411764705882, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5882352941176471, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.17647058823529413, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.43509615384615385, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.45290396735072136, + "epoch": 0.06535947712418301, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.7938478589057922, + "learning_rate": 9.889612861977853e-07, + "loss": 0.0549, + "reward": 1.0546875, + "reward_std": 0.4169638752937317, + "rewards/MultiTurnTaskReward/mean": 1.0546875, + "rewards/MultiTurnTaskReward/std": 0.4989531338214874, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0004034042358398, + "sampling/importance_sampling_ratio/min": 0.6148865222930908, + "sampling/sampling_logp_difference/max": 0.769619345664978, + "sampling/sampling_logp_difference/mean": 0.015473594889044762, + "step": 20 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9184027777777778, + "aime/avg_subtask_length": 136.89987163029525, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.005671077504725898, + "2": 0.13610586011342155, + "3": 0.77882797731569, + "4": 0.0661625708884688, + "5": 0.013232514177693762 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.08159722222222222, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.010911424903722721, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.22913992297817715, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.02053915275994865, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7503209242618742, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.045368620037807186, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9546313799621928, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.34790874524714827, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.013307984790874524, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6387832699619772, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.30176211453744495, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.05066079295154185, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6475770925110133, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2857142857142857, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.047619047619047616, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6666666666666666, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.14285714285714285, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8571428571428571, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7743055555555556, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0078125, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 693.0, + "completions/mean_length": 302.2890625, + "completions/mean_terminated_length": 296.6062927246094, + "completions/min_length": 101.0, + "completions/min_terminated_length": 101.0, + "epoch": 0.06535947712418301, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8080357142857143, + "global/avg_subtask_length": 125.14961679440187, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07366482504604052, + "2": 0.19152854511970535, + "3": 0.6546961325966851, + "4": 0.058011049723756904, + "5": 0.022099447513812154 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.19196428571428573, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.06464511829390203, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.32655781406197937, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1732755748083972, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.5001666111296235, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3397790055248619, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.1887661141804788, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4714548802946593, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.33697813121272363, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1242544731610338, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5387673956262425, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.29949874686716793, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.19799498746867167, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5025062656641605, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3218390804597701, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.25287356321839083, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.42528735632183906, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.20833333333333334, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4166666666666667, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.375, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.5907738095238095, + "global/total_cost": 0.0, + "num_tokens": 5955559.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8181818181818182, + "simpleqa_verified/avg_subtask_length": 107.73544093178036, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2604166666666667, + "2": 0.3993055555555556, + "3": 0.3333333333333333, + "4": 0.006944444444444444, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.18181818181818182, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.10149750415973377, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.632279534109817, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.064891846921797, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.30282861896838603, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9895833333333334, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.010416666666666666, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.18779342723004694, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.06103286384976526, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7511737089201878, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5408163265306123, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2653061224489796, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.19387755102040816, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.4744318181818182, + "simpleqa_verified/total_cost": 0.0, + "step": 20, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.6466346153846154, + "swe_bench/avg_subtask_length": 115.8372921615202, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.007434944237918215, + "2": 0.07806691449814127, + "3": 0.7546468401486989, + "4": 0.09665427509293681, + "5": 0.06319702602230483 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.35336538461538464, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1377672209026128, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.28859857482185275, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5332541567695962, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17814726840855108, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22304832713754646, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7620817843866171, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.01486988847583643, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4344569288389513, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.39325842696629215, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.17228464419475656, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.1991869918699187, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.44308943089430897, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.35772357723577236, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.32558139534883723, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46511627906976744, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20930232558139536, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.23529411764705882, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5882352941176471, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.17647058823529413, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.43509615384615385, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.37303343042731285, + "epoch": 0.06862745098039216, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.8787543177604675, + "learning_rate": 9.872052623234631e-07, + "loss": 0.0798, + "reward": 1.31640625, + "reward_std": 0.43004271388053894, + "rewards/MultiTurnTaskReward/mean": 1.31640625, + "rewards/MultiTurnTaskReward/std": 0.45190364122390747, + "sampling/importance_sampling_ratio/max": 1.6531881093978882, + "sampling/importance_sampling_ratio/mean": 1.0000722408294678, + "sampling/importance_sampling_ratio/min": 0.5480276942253113, + "sampling/sampling_logp_difference/max": 0.6014294624328613, + "sampling/sampling_logp_difference/mean": 0.014696512371301651, + "step": 21 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9184027777777778, + "aime/avg_subtask_length": 136.89987163029525, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.005671077504725898, + "2": 0.13610586011342155, + "3": 0.77882797731569, + "4": 0.0661625708884688, + "5": 0.013232514177693762 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.08159722222222222, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.010911424903722721, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.22913992297817715, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.02053915275994865, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7503209242618742, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.045368620037807186, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9546313799621928, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.34790874524714827, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.013307984790874524, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6387832699619772, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.30176211453744495, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.05066079295154185, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6475770925110133, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2857142857142857, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.047619047619047616, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6666666666666666, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.14285714285714285, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8571428571428571, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7743055555555556, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 702.0, + "completions/max_terminated_length": 702.0, + "completions/mean_length": 286.3046875, + "completions/mean_terminated_length": 286.3046875, + "completions/min_length": 80.0, + "completions/min_terminated_length": 80.0, + "epoch": 0.06862745098039216, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8167613636363636, + "global/avg_subtask_length": 124.74063581995593, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07391304347826087, + "2": 0.19391304347826088, + "3": 0.6513043478260869, + "4": 0.057391304347826085, + "5": 0.023478260869565216 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.18323863636363635, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.06106389675794775, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3320742839156437, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18287692791942084, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.48504878816493546, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.35478260869565215, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.44521739130434784, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.336150234741784, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1323943661971831, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5314553990610329, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3028503562945368, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2042755344418052, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.49287410926365793, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3118279569892473, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.26881720430107525, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.41935483870967744, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18518518518518517, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.48148148148148145, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3333333333333333, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6065340909090909, + "global/total_cost": 0.0, + "num_tokens": 6222094.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8333333333333334, + "simpleqa_verified/avg_subtask_length": 108.45994065281899, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.25, + "2": 0.4, + "3": 0.34375, + "4": 0.00625, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.16666666666666666, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.09050445103857567, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6335311572700296, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.06231454005934718, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.30415430267062316, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.990625, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.009375, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.19166666666666668, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.058333333333333334, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.75, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5535714285714286, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.25, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.19642857142857142, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5104166666666666, + "simpleqa_verified/total_cost": 0.0, + "step": 21, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.671875, + "swe_bench/avg_subtask_length": 116.3058201058201, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.006644518272425249, + "2": 0.07641196013289037, + "3": 0.7541528239202658, + "4": 0.09634551495016612, + "5": 0.0664451827242525 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.328125, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.12275132275132275, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2867724867724868, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5365079365079365, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17671957671957672, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22259136212624583, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7641196013289037, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.013289036544850499, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.431438127090301, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4013377926421405, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.16722408026755853, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2028985507246377, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4384057971014493, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.358695652173913, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30612244897959184, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46938775510204084, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.22448979591836735, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.65, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.15, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.4732142857142857, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3970301579684019, + "epoch": 0.0718954248366013, + "frac_reward_zero_std": 0.25, + "grad_norm": 0.7542834877967834, + "learning_rate": 9.853214773129795e-07, + "loss": -0.0572, + "reward": 1.41796875, + "reward_std": 0.24061989784240723, + "rewards/MultiTurnTaskReward/mean": 1.41796875, + "rewards/MultiTurnTaskReward/std": 0.30596011877059937, + "sampling/importance_sampling_ratio/max": 1.9899441003799438, + "sampling/importance_sampling_ratio/mean": 1.0003106594085693, + "sampling/importance_sampling_ratio/min": 0.34070512652397156, + "sampling/sampling_logp_difference/max": 1.076737880706787, + "sampling/sampling_logp_difference/mean": 0.01564490608870983, + "step": 22 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9226973684210527, + "aime/avg_subtask_length": 136.32907542579076, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0071301247771836, + "2": 0.14260249554367202, + "3": 0.7754010695187166, + "4": 0.062388591800356503, + "5": 0.012477718360071301 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.07730263157894737, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.010340632603406326, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.23357664233576642, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.021897810218978103, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7445255474452555, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.0427807486631016, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9572192513368984, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.36804308797127466, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.012567324955116697, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6193895870736086, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2976939203354298, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.05660377358490566, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6457023060796646, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2857142857142857, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.047619047619047616, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6666666666666666, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.14285714285714285, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8571428571428571, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.78125, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 690.0, + "completions/max_terminated_length": 690.0, + "completions/mean_length": 257.3671875, + "completions/mean_terminated_length": 257.3671875, + "completions/min_length": 106.0, + "completions/min_terminated_length": 106.0, + "epoch": 0.0718954248366013, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8247282608695652, + "global/avg_subtask_length": 124.51710684273709, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07825370675453047, + "2": 0.19769357495881384, + "3": 0.6474464579901154, + "4": 0.054365733113673806, + "5": 0.022240527182866558 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.17527173913043478, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.05822328931572629, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.33913565426170467, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.17677070828331332, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.484093637454982, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3624382207578254, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.18945634266886327, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.44810543657331137, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3449508489722967, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1260053619302949, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5290437890974083, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.30716723549488056, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.20477815699658702, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4880546075085324, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3118279569892473, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.26881720430107525, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.41935483870967744, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18518518518518517, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.48148148148148145, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3333333333333333, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6195652173913043, + "global/total_cost": 0.0, + "num_tokens": 6486333.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8461538461538461, + "simpleqa_verified/avg_subtask_length": 108.8250336473755, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2528409090909091, + "2": 0.38920454545454547, + "3": 0.3522727272727273, + "4": 0.005681818181818182, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.15384615384615385, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.08209959623149395, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.639300134589502, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.06191117092866756, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.29878869448183043, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9914772727272727, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.008522727272727272, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.19771863117870722, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.053231939163498096, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7490494296577946, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5714285714285714, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.25396825396825395, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.1746031746031746, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5408653846153846, + "simpleqa_verified/total_cost": 0.0, + "step": 22, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.671875, + "swe_bench/avg_subtask_length": 116.3058201058201, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.006644518272425249, + "2": 0.07641196013289037, + "3": 0.7541528239202658, + "4": 0.09634551495016612, + "5": 0.0664451827242525 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.328125, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.12275132275132275, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2867724867724868, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5365079365079365, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17671957671957672, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22259136212624583, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7641196013289037, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.013289036544850499, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.431438127090301, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4013377926421405, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.16722408026755853, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2028985507246377, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4384057971014493, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.358695652173913, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30612244897959184, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46938775510204084, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.22448979591836735, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.65, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.15, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.4732142857142857, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4187340922653675, + "epoch": 0.07516339869281045, + "frac_reward_zero_std": 0.25, + "grad_norm": 0.9385222792625427, + "learning_rate": 9.833104251563055e-07, + "loss": 0.0732, + "reward": 1.4296875, + "reward_std": 0.24029779434204102, + "rewards/MultiTurnTaskReward/mean": 1.4296875, + "rewards/MultiTurnTaskReward/std": 0.2787371277809143, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9998148679733276, + "sampling/importance_sampling_ratio/min": 0.43095114827156067, + "sampling/sampling_logp_difference/max": 0.841760516166687, + "sampling/sampling_logp_difference/mean": 0.016574088484048843, + "step": 23 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9226973684210527, + "aime/avg_subtask_length": 136.32907542579076, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0071301247771836, + "2": 0.14260249554367202, + "3": 0.7754010695187166, + "4": 0.062388591800356503, + "5": 0.012477718360071301 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.07730263157894737, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.010340632603406326, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.23357664233576642, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.021897810218978103, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7445255474452555, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.0427807486631016, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9572192513368984, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.36804308797127466, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.012567324955116697, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6193895870736086, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2976939203354298, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.05660377358490566, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6457023060796646, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2857142857142857, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.047619047619047616, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6666666666666666, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.14285714285714285, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8571428571428571, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.78125, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 779.0, + "completions/max_terminated_length": 779.0, + "completions/mean_length": 220.5, + "completions/mean_terminated_length": 220.5, + "completions/min_length": 68.0, + "completions/min_terminated_length": 68.0, + "epoch": 0.07516339869281045, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8313802083333334, + "global/avg_subtask_length": 123.50300601202404, + "global/chosen_num_routing_steps_ratio": { + "1": 0.08300704776820673, + "2": 0.19498825371965545, + "3": 0.6476115896632733, + "4": 0.052466718872357085, + "5": 0.02192638997650744 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.16861979166666666, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.05553965073003149, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.34297165760091614, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1837961637560836, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.47323217864300027, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3766640563821457, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.19733750978856696, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.42599843382928737, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.34244235695986336, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1340734415029889, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5234842015371477, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.30585683297180044, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.20932754880694143, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4848156182212581, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30526315789473685, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2736842105263158, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.42105263157894735, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.17857142857142858, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.32142857142857145, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6139322916666666, + "global/total_cost": 0.0, + "num_tokens": 6733085.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8549107142857143, + "simpleqa_verified/avg_subtask_length": 105.90942928039702, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.26109660574412535, + "2": 0.3785900783289817, + "3": 0.35509138381201044, + "4": 0.005221932114882507, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.14508928571428573, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.07568238213399504, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6401985111662531, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.062034739454094295, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.2977667493796526, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9921671018276762, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007832898172323759, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.19081272084805653, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.05653710247349823, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7526501766784452, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5797101449275363, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2463768115942029, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.17391304347826086, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5714285714285714, + "simpleqa_verified/total_cost": 0.0, + "step": 23, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.69375, + "swe_bench/avg_subtask_length": 116.88207094918505, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.006006006006006006, + "2": 0.07207207207207207, + "3": 0.7687687687687688, + "4": 0.09009009009009009, + "5": 0.06306306306306306 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.30625, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.1112176414189837, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2857142857142857, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5330776605944392, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.18120805369127516, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.23123123123123124, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7567567567567568, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012012012012012012, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.42900302114803623, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.40483383685800606, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.1661631419939577, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.19543973941368079, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.42996742671009774, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3745928338762215, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.29411764705882354, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.47058823529411764, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.23529411764705882, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.19047619047619047, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6666666666666666, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.14285714285714285, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.44166666666666665, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.40351391583681107, + "epoch": 0.0784313725490196, + "frac_reward_zero_std": 0.25, + "grad_norm": 0.892056941986084, + "learning_rate": 9.81172633217015e-07, + "loss": -0.0319, + "reward": 0.73828125, + "reward_std": 0.13843242824077606, + "rewards/MultiTurnTaskReward/mean": 0.73828125, + "rewards/MultiTurnTaskReward/std": 0.44793859124183655, + "sampling/importance_sampling_ratio/max": 1.6169949769973755, + "sampling/importance_sampling_ratio/mean": 0.9999903440475464, + "sampling/importance_sampling_ratio/min": 0.4775344133377075, + "sampling/sampling_logp_difference/max": 0.7391190528869629, + "sampling/sampling_logp_difference/mean": 0.017379477620124817, + "step": 24 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9226973684210527, + "aime/avg_subtask_length": 136.32907542579076, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0071301247771836, + "2": 0.14260249554367202, + "3": 0.7754010695187166, + "4": 0.062388591800356503, + "5": 0.012477718360071301 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.07730263157894737, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.010340632603406326, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.23357664233576642, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.021897810218978103, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7445255474452555, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.0427807486631016, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9572192513368984, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.36804308797127466, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.012567324955116697, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6193895870736086, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2976939203354298, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.05660377358490566, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6457023060796646, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2857142857142857, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.047619047619047616, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6666666666666666, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.14285714285714285, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8571428571428571, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.78125, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 951.0, + "completions/max_terminated_length": 951.0, + "completions/mean_length": 278.015625, + "completions/mean_terminated_length": 278.015625, + "completions/min_length": 96.0, + "completions/min_terminated_length": 96.0, + "epoch": 0.0784313725490196, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.838125, + "global/avg_subtask_length": 124.05783783783784, + "global/chosen_num_routing_steps_ratio": { + "1": 0.0790454884414616, + "2": 0.18791946308724833, + "3": 0.6524981357196122, + "4": 0.05592841163310962, + "5": 0.024608501118568233 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.161875, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.05243243243243243, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3418918918918919, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.20324324324324325, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4548648648648649, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3698732289336316, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22371364653243847, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4064131245339299, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.35465587044534413, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1465587044534413, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4987854251012146, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2950152594099695, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2268565615462869, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.47812817904374366, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3148148148148148, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2962962962962963, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3888888888888889, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.21212121212121213, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.48484848484848486, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.30303030303030304, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.629375, + "global/total_cost": 0.0, + "num_tokens": 7068479.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8549107142857143, + "simpleqa_verified/avg_subtask_length": 105.90942928039702, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.26109660574412535, + "2": 0.3785900783289817, + "3": 0.35509138381201044, + "4": 0.005221932114882507, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.14508928571428573, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.07568238213399504, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6401985111662531, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.062034739454094295, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.2977667493796526, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9921671018276762, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007832898172323759, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.19081272084805653, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.05653710247349823, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7526501766784452, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5797101449275363, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2463768115942029, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.17391304347826086, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5714285714285714, + "simpleqa_verified/total_cost": 0.0, + "step": 24, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.7297794117647058, + "swe_bench/avg_subtask_length": 119.6208, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.005037783375314861, + "2": 0.06801007556675064, + "3": 0.7657430730478589, + "4": 0.09571788413098237, + "5": 0.0654911838790932 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.2702205882352941, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.0928, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.292, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5328, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1752, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.23173803526448364, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7556675062972292, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012594458438287154, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4531645569620253, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.1468354430379747, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18478260869565216, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.44021739130434784, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.375, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3125, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46875, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.21875, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.23076923076923078, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6153846153846154, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.15384615384615385, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5073529411764706, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3809783924371004, + "epoch": 0.08169934640522876, + "frac_reward_zero_std": 0.5, + "grad_norm": 0.7983564734458923, + "learning_rate": 9.789086620939935e-07, + "loss": 0.0831, + "reward": 1.44921875, + "reward_std": 0.17366787791252136, + "rewards/MultiTurnTaskReward/mean": 1.44921875, + "rewards/MultiTurnTaskReward/std": 0.25748515129089355, + "sampling/importance_sampling_ratio/max": 1.9394257068634033, + "sampling/importance_sampling_ratio/mean": 0.9999889731407166, + "sampling/importance_sampling_ratio/min": 0.439921498298645, + "sampling/sampling_logp_difference/max": 0.8211590051651001, + "sampling/sampling_logp_difference/mean": 0.018395788967609406, + "step": 25 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.925, + "aime/avg_subtask_length": 136.93904542840713, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.006756756756756757, + "2": 0.14020270270270271, + "3": 0.777027027027027, + "4": 0.060810810810810814, + "5": 0.015202702702702704 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.075, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0097757331799885, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.23461759631972398, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.02127659574468085, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7441058079355952, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.04560810810810811, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9543918918918919, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3741496598639456, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011904761904761904, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6139455782312925, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.29306930693069305, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.053465346534653464, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6534653465346535, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.26666666666666666, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.06666666666666667, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6666666666666666, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1111111111111111, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8888888888888888, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.790625, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 637.0, + "completions/max_terminated_length": 637.0, + "completions/mean_length": 288.2734375, + "completions/mean_terminated_length": 288.2734375, + "completions/min_length": 112.0, + "completions/min_terminated_length": 112.0, + "epoch": 0.08169934640522876, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8431490384615384, + "global/avg_subtask_length": 125.00951156812339, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07555238774055595, + "2": 0.18246614397719174, + "3": 0.6607270135424091, + "4": 0.05630791161796151, + "5": 0.024946543121881683 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.15685096153846154, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.04987146529562982, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3390745501285347, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.20488431876606683, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.45604113110539846, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.35994297933000713, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.23093371347113328, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4091233071988596, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3646877409406322, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.14649190439475712, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.48882035466461066, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2881844380403458, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2219020172910663, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4899135446685879, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2982456140350877, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3157894736842105, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.38596491228070173, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.45714285714285713, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.34285714285714286, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6424278846153846, + "global/total_cost": 0.0, + "num_tokens": 7418338.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8549107142857143, + "simpleqa_verified/avg_subtask_length": 105.90942928039702, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.26109660574412535, + "2": 0.3785900783289817, + "3": 0.35509138381201044, + "4": 0.005221932114882507, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.14508928571428573, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.07568238213399504, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6401985111662531, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.062034739454094295, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.2977667493796526, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9921671018276762, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007832898172323759, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.19081272084805653, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.05653710247349823, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7526501766784452, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5797101449275363, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2463768115942029, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.17391304347826086, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5714285714285714, + "simpleqa_verified/total_cost": 0.0, + "step": 25, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.7430555555555556, + "swe_bench/avg_subtask_length": 121.03122676579926, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.004672897196261682, + "2": 0.06542056074766354, + "3": 0.7733644859813084, + "4": 0.09579439252336448, + "5": 0.06074766355140187 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.2569444444444444, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.0862453531598513, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2936802973977695, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5278810408921933, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17843866171003717, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22897196261682243, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7570093457943925, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.014018691588785047, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4671361502347418, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.392018779342723, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.14084507042253522, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18090452261306533, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4271356783919598, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.39195979899497485, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.29850746268656714, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4925373134328358, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.208955223880597, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.23076923076923078, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6153846153846154, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.15384615384615385, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5329861111111112, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4238539803773165, + "epoch": 0.08496732026143791, + "frac_reward_zero_std": 0.25, + "grad_norm": 0.9228416085243225, + "learning_rate": 9.765191054744304e-07, + "loss": 0.0393, + "reward": 1.46484375, + "reward_std": 0.19887377321720123, + "rewards/MultiTurnTaskReward/mean": 1.46484375, + "rewards/MultiTurnTaskReward/std": 0.22782430052757263, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9999989867210388, + "sampling/importance_sampling_ratio/min": 0.4309717118740082, + "sampling/sampling_logp_difference/max": 0.8417128324508667, + "sampling/sampling_logp_difference/mean": 0.0184822715818882, + "step": 26 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9270833333333334, + "aime/avg_subtask_length": 136.8146528157463, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.006420545746388443, + "2": 0.14285714285714285, + "3": 0.7736757624398074, + "4": 0.06260032102728733, + "5": 0.014446227929373997 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.07291666666666667, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.009294696555494806, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 2, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2378348824494259, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.021323127392017496, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7408419901585566, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.0449438202247191, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9550561797752809, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.38449111470113084, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.012924071082390954, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6025848142164781, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.29245283018867924, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.0509433962264151, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6566037735849056, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2708333333333333, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.08333333333333333, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6458333333333334, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1111111111111111, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8888888888888888, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7931547619047619, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 550.0, + "completions/max_terminated_length": 550.0, + "completions/mean_length": 233.375, + "completions/mean_terminated_length": 233.375, + "completions/min_length": 107.0, + "completions/min_terminated_length": 107.0, + "epoch": 0.08496732026143791, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8483796296296297, + "global/avg_subtask_length": 124.64351050679852, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07844474761255116, + "2": 0.18758526603001363, + "3": 0.6541609822646658, + "4": 0.05593451568894952, + "5": 0.023874488403819918 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.15162037037037038, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.04796044499381953, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 2, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.34313967861557476, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19826946847960444, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.45859085290482077, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3669849931787176, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22100954979536153, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4120054570259209, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.36565507031828276, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.14285714285714285, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4914877868245744, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.29182156133828996, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21561338289962825, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.49256505576208176, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.29914529914529914, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3162393162393162, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.38461538461538464, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.45714285714285713, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.34285714285714286, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6342592592592593, + "global/total_cost": 0.0, + "num_tokens": 7627058.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8645833333333334, + "simpleqa_verified/avg_subtask_length": 104.66360505166476, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.26265060240963856, + "2": 0.38072289156626504, + "3": 0.35180722891566263, + "4": 0.004819277108433735, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.13541666666666666, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.07003444316877153, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6406429391504018, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.060849598163030996, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.29850746268656714, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9927710843373494, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007228915662650603, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.18627450980392157, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.058823529411764705, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7549019607843137, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5878378378378378, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.23648648648648649, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.17567567567567569, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5333333333333333, + "simpleqa_verified/total_cost": 0.0, + "step": 26, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.7430555555555556, + "swe_bench/avg_subtask_length": 121.03122676579926, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.004672897196261682, + "2": 0.06542056074766354, + "3": 0.7733644859813084, + "4": 0.09579439252336448, + "5": 0.06074766355140187 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.2569444444444444, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.0862453531598513, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2936802973977695, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5278810408921933, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17843866171003717, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22897196261682243, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7570093457943925, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.014018691588785047, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4671361502347418, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.392018779342723, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.14084507042253522, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18090452261306533, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4271356783919598, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.39195979899497485, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.29850746268656714, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4925373134328358, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.208955223880597, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.23076923076923078, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6153846153846154, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.15384615384615385, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5329861111111112, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4149255231022835, + "epoch": 0.08823529411764706, + "frac_reward_zero_std": 0.25, + "grad_norm": 0.8541430830955505, + "learning_rate": 9.740045899781352e-07, + "loss": -0.0595, + "reward": 1.18359375, + "reward_std": 0.20909416675567627, + "rewards/MultiTurnTaskReward/mean": 1.18359375, + "rewards/MultiTurnTaskReward/std": 0.4731827676296234, + "sampling/importance_sampling_ratio/max": 1.6067310571670532, + "sampling/importance_sampling_ratio/mean": 1.000626802444458, + "sampling/importance_sampling_ratio/min": 0.2349558174610138, + "sampling/sampling_logp_difference/max": 1.4483578205108643, + "sampling/sampling_logp_difference/mean": 0.016654271632432938, + "step": 27 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9303977272727273, + "aime/avg_subtask_length": 136.74272349272348, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0061068702290076335, + "2": 0.14045801526717558, + "3": 0.7770992366412214, + "4": 0.06259541984732825, + "5": 0.013740458015267175 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.06960227272727272, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.008835758835758836, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 4, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2396049896049896, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.02027027027027027, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7401247401247402, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.04580152671755725, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9541984732824428, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.38556067588325654, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01228878648233487, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6021505376344086, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.295169946332737, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04830053667262969, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6565295169946332, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.28, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.08, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.64, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1111111111111111, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8888888888888888, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7954545454545454, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 551.0, + "completions/max_terminated_length": 551.0, + "completions/mean_length": 267.328125, + "completions/mean_terminated_length": 267.328125, + "completions/min_length": 120.0, + "completions/min_terminated_length": 120.0, + "epoch": 0.08823529411764706, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8526785714285714, + "global/avg_subtask_length": 125.20373434176318, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07526178010471204, + "2": 0.18324607329842932, + "3": 0.6616492146596858, + "4": 0.05693717277486911, + "5": 0.022905759162303665 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.14732142857142858, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.04585204443393997, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 4, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3415268258095013, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.20042543134010873, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.45804774285038996, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.35471204188481675, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.23036649214659685, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.41492146596858637, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3750884642604388, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.14012738853503184, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4847841472045294, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.28949691085613416, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21535745807590467, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.49514563106796117, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3114754098360656, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3114754098360656, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3770491803278688, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.45714285714285713, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.34285714285714286, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6428571428571429, + "global/total_cost": 0.0, + "num_tokens": 7916700.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8645833333333334, + "simpleqa_verified/avg_subtask_length": 104.66360505166476, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.26265060240963856, + "2": 0.38072289156626504, + "3": 0.35180722891566263, + "4": 0.004819277108433735, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.13541666666666666, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.07003444316877153, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6406429391504018, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.060849598163030996, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.29850746268656714, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9927710843373494, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007228915662650603, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.18627450980392157, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.058823529411764705, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7549019607843137, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5878378378378378, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.23648648648648649, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.17567567567567569, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5333333333333333, + "simpleqa_verified/total_cost": 0.0, + "step": 27, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.7532894736842105, + "swe_bench/avg_subtask_length": 122.20194986072423, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.004366812227074236, + "2": 0.06550218340611354, + "3": 0.777292576419214, + "4": 0.09606986899563319, + "5": 0.056768558951965066 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.24671052631578946, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.0807799442896936, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2966573816155989, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5264623955431755, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17688022284122562, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2183406113537118, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7685589519650655, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.013100436681222707, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4868421052631579, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.37719298245614036, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13596491228070176, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.1784037558685446, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4272300469483568, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.39436619718309857, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3142857142857143, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4857142857142857, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.23076923076923078, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6153846153846154, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.15384615384615385, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5526315789473685, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.42329093627631664, + "epoch": 0.0915032679738562, + "frac_reward_zero_std": 0.0, + "grad_norm": 0.8756455779075623, + "learning_rate": 9.713657749932171e-07, + "loss": 0.0216, + "reward": 1.171875, + "reward_std": 0.34633708000183105, + "rewards/MultiTurnTaskReward/mean": 1.171875, + "rewards/MultiTurnTaskReward/std": 0.4957992732524872, + "sampling/importance_sampling_ratio/max": 1.8649109601974487, + "sampling/importance_sampling_ratio/mean": 0.9999725818634033, + "sampling/importance_sampling_ratio/min": 0.28228241205215454, + "sampling/sampling_logp_difference/max": 1.2648471593856812, + "sampling/sampling_logp_difference/mean": 0.01794636994600296, + "step": 28 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9334239130434783, + "aime/avg_subtask_length": 136.76701440635867, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.005822416302765648, + "2": 0.14556040756914118, + "3": 0.7743813682678311, + "4": 0.0611353711790393, + "5": 0.013100436681222707 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.06657608695652174, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.008445106805762544, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 4, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2429210134128167, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.020864381520119227, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.736214605067064, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.04512372634643377, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9548762736535662, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3923865300146413, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01171303074670571, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.595900439238653, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.30017152658662094, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.051457975986277875, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6483704974271012, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.27450980392156865, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0784313725490196, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6470588235294118, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1111111111111111, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8888888888888888, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7975543478260869, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 683.0, + "completions/max_terminated_length": 683.0, + "completions/mean_length": 254.625, + "completions/mean_terminated_length": 254.625, + "completions/min_length": 83.0, + "completions/min_terminated_length": 83.0, + "epoch": 0.0915032679738562, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8577586206896551, + "global/avg_subtask_length": 125.07687074829931, + "global/chosen_num_routing_steps_ratio": { + "1": 0.0722361809045226, + "2": 0.18530150753768845, + "3": 0.664572864321608, + "4": 0.055904522613065326, + "5": 0.02198492462311558 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.14224137931034483, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.04399092970521542, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 4, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3401360544217687, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.20272108843537415, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.45714285714285713, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3473618090452261, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2342964824120603, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4183417085427136, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3784698713608666, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.14353419092755584, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4779959377115775, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.29018612521150594, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21573604060913706, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.494077834179357, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3064516129032258, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3064516129032258, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3870967741935484, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.45714285714285713, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.34285714285714286, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6357758620689655, + "global/total_cost": 0.0, + "num_tokens": 8264748.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8645833333333334, + "simpleqa_verified/avg_subtask_length": 104.66360505166476, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.26265060240963856, + "2": 0.38072289156626504, + "3": 0.35180722891566263, + "4": 0.004819277108433735, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.13541666666666666, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.07003444316877153, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6406429391504018, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.060849598163030996, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.29850746268656714, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9927710843373494, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007228915662650603, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.18627450980392157, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.058823529411764705, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7549019607843137, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5878378378378378, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.23648648648648649, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.17567567567567569, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5333333333333333, + "simpleqa_verified/total_cost": 0.0, + "step": 28, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.765625, + "swe_bench/avg_subtask_length": 121.30733944954129, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.004081632653061225, + "2": 0.07551020408163266, + "3": 0.7755102040816326, + "4": 0.09183673469387756, + "5": 0.053061224489795916 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.234375, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.07601572739187418, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.29685452162516385, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5235910878112713, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17955439056356487, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22448979591836735, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7612244897959184, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.014285714285714285, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.47950819672131145, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.38114754098360654, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13934426229508196, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.17960088691796008, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4212860310421286, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3991130820399113, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30985915492957744, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4788732394366197, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2112676056338028, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.23076923076923078, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6153846153846154, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.15384615384615385, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5265625, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.40079533122479916, + "epoch": 0.09477124183006536, + "frac_reward_zero_std": 0.25, + "grad_norm": 0.9488219022750854, + "learning_rate": 9.686033525031719e-07, + "loss": 0.0095, + "reward": 1.19921875, + "reward_std": 0.21489672362804413, + "rewards/MultiTurnTaskReward/mean": 1.19921875, + "rewards/MultiTurnTaskReward/std": 0.46676909923553467, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9997531175613403, + "sampling/importance_sampling_ratio/min": 0.5041602253913879, + "sampling/sampling_logp_difference/max": 1.1879768371582031, + "sampling/sampling_logp_difference/mean": 0.018169153481721878, + "step": 29 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9334239130434783, + "aime/avg_subtask_length": 136.76701440635867, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.005822416302765648, + "2": 0.14556040756914118, + "3": 0.7743813682678311, + "4": 0.0611353711790393, + "5": 0.013100436681222707 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.06657608695652174, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.008445106805762544, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 4, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2429210134128167, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.020864381520119227, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.736214605067064, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.04512372634643377, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9548762736535662, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3923865300146413, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01171303074670571, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.595900439238653, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.30017152658662094, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.051457975986277875, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6483704974271012, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.27450980392156865, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0784313725490196, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6470588235294118, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1111111111111111, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8888888888888888, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.7975543478260869, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 513.0, + "completions/max_terminated_length": 513.0, + "completions/mean_length": 217.1953125, + "completions/mean_terminated_length": 217.1953125, + "completions/min_length": 66.0, + "completions/min_terminated_length": 66.0, + "epoch": 0.09477124183006536, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8625, + "global/avg_subtask_length": 124.00462249614792, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07910628019323672, + "2": 0.19444444444444445, + "3": 0.6515700483091788, + "4": 0.05374396135265701, + "5": 0.021135265700483092 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.1375, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.042703059652212196, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 4, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.348448162007484, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19854721549636803, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4530046224961479, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3719806763285024, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2252415458937198, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4027777777777778, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3718032786885246, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.14032786885245901, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4878688524590164, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2950955943474647, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2169576059850374, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.48794679966749793, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3064516129032258, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3064516129032258, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3870967741935484, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.45714285714285713, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.34285714285714286, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6390625, + "global/total_cost": 0.0, + "num_tokens": 8524837.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8805147058823529, + "simpleqa_verified/avg_subtask_length": 102.51593625498008, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2609603340292276, + "2": 0.3862212943632568, + "3": 0.348643006263048, + "4": 0.0041753653444676405, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.11948529411764706, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.060756972111553786, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6384462151394422, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.060756972111553786, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.300796812749004, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9916492693110647, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.008350730688935281, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.18361581920903955, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.05649717514124294, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7598870056497176, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5857988165680473, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.24260355029585798, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.17159763313609466, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5569852941176471, + "simpleqa_verified/total_cost": 0.0, + "step": 29, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.765625, + "swe_bench/avg_subtask_length": 121.30733944954129, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.004081632653061225, + "2": 0.07551020408163266, + "3": 0.7755102040816326, + "4": 0.09183673469387756, + "5": 0.053061224489795916 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.234375, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.07601572739187418, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.29685452162516385, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5235910878112713, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17955439056356487, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22448979591836735, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7612244897959184, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.014285714285714285, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.47950819672131145, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.38114754098360654, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13934426229508196, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.17960088691796008, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4212860310421286, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3991130820399113, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30985915492957744, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4788732394366197, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2112676056338028, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.23076923076923078, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6153846153846154, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.15384615384615385, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5265625, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.39473845809698105, + "epoch": 0.09803921568627451, + "frac_reward_zero_std": 0.5, + "grad_norm": 0.8809399604797363, + "learning_rate": 9.657180469054212e-07, + "loss": 0.0268, + "reward": 1.328125, + "reward_std": 0.21897734701633453, + "rewards/MultiTurnTaskReward/mean": 1.328125, + "rewards/MultiTurnTaskReward/std": 0.3787541687488556, + "sampling/importance_sampling_ratio/max": 1.560965657234192, + "sampling/importance_sampling_ratio/mean": 0.9999340772628784, + "sampling/importance_sampling_ratio/min": 0.29176244139671326, + "sampling/sampling_logp_difference/max": 1.2318153381347656, + "sampling/sampling_logp_difference/mean": 0.017089931294322014, + "step": 30 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9361979166666666, + "aime/avg_subtask_length": 136.7797252486973, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.005563282336578581, + "2": 0.14186369958275383, + "3": 0.7788595271210014, + "4": 0.0584144645340751, + "5": 0.015299026425591099 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.06380208333333333, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.008053055423969682, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.24585504500236854, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.02036949313121743, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7337754618664141, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.043115438108484005, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9568845618915159, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3958041958041958, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011188811188811189, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.593006993006993, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3066884176182708, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.05057096247960848, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6427406199021207, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2830188679245283, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.07547169811320754, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6415094339622641, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8181818181818182, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8033854166666666, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 611.0, + "completions/max_terminated_length": 611.0, + "completions/mean_length": 229.984375, + "completions/mean_terminated_length": 229.984375, + "completions/min_length": 106.0, + "completions/min_terminated_length": 106.0, + "epoch": 0.09803921568627451, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8669354838709677, + "global/avg_subtask_length": 124.39593047901653, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07732558139534884, + "2": 0.19709302325581396, + "3": 0.6523255813953488, + "4": 0.051744186046511625, + "5": 0.021511627906976746 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.13306451612903225, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.04111911827045358, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.35184400169563373, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.192030521407376, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.45612547689699023, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3755813953488372, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21686046511627907, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.40755813953488373, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.37177063642091995, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1348456206679269, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4933837429111531, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3020833333333333, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21233974358974358, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4855769230769231, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30952380952380953, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.30158730158730157, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3888888888888889, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.21621621621621623, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.43243243243243246, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.35135135135135137, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6496975806451613, + "global/total_cost": 0.0, + "num_tokens": 8772579.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8871527777777778, + "simpleqa_verified/avg_subtask_length": 104.57261794634597, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.24853228962818003, + "2": 0.3913894324853229, + "3": 0.3561643835616438, + "4": 0.003913894324853229, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.11284722222222222, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.056429232192414434, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6364477335800185, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.05920444033302498, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.30434782608695654, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9882583170254403, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011741682974559686, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.19010416666666666, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.052083333333333336, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7578125, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5869565217391305, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2391304347826087, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.17391304347826086, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5815972222222222, + "simpleqa_verified/total_cost": 0.0, + "step": 30, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.765625, + "swe_bench/avg_subtask_length": 121.30733944954129, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.004081632653061225, + "2": 0.07551020408163266, + "3": 0.7755102040816326, + "4": 0.09183673469387756, + "5": 0.053061224489795916 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.234375, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.07601572739187418, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.29685452162516385, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5235910878112713, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17955439056356487, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22448979591836735, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7612244897959184, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.014285714285714285, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.47950819672131145, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.38114754098360654, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13934426229508196, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.17960088691796008, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4212860310421286, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3991130820399113, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30985915492957744, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4788732394366197, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2112676056338028, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.23076923076923078, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6153846153846154, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.15384615384615385, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5265625, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.38000036031007767, + "epoch": 0.10130718954248366, + "frac_reward_zero_std": 0.5, + "grad_norm": 0.8584945201873779, + "learning_rate": 9.62710614821352e-07, + "loss": 0.0186, + "reward": 1.46875, + "reward_std": 0.12296734005212784, + "rewards/MultiTurnTaskReward/mean": 1.46875, + "rewards/MultiTurnTaskReward/std": 0.1746762990951538, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0001206398010254, + "sampling/importance_sampling_ratio/min": 0.44446563720703125, + "sampling/sampling_logp_difference/max": 0.9451560974121094, + "sampling/sampling_logp_difference/mean": 0.017514154314994812, + "step": 31 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9362980769230769, + "aime/avg_subtask_length": 137.5522518583297, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.005134788189987163, + "2": 0.14249037227214378, + "3": 0.7779204107830552, + "4": 0.06033376123234917, + "5": 0.014120667522464698 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.06370192307692307, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.007433318758198513, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.24617402710975075, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.020113686051595976, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7337122868386533, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03979460847240052, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9602053915275995, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3961290322580645, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011612903225806452, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5922580645161291, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3117469879518072, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04819277108433735, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6400602409638554, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.27586206896551724, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.08620689655172414, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6379310344827587, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8181818181818182, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8100961538461539, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.015625, + "completions/max_length": 1024.0, + "completions/max_terminated_length": 483.0, + "completions/mean_length": 243.015625, + "completions/mean_terminated_length": 230.6190643310547, + "completions/min_length": 92.0, + "completions/min_terminated_length": 92.0, + "epoch": 0.10130718954248366, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.869140625, + "global/avg_subtask_length": 125.20228851655088, + "global/chosen_num_routing_steps_ratio": { + "1": 0.0747191011235955, + "2": 0.19550561797752808, + "3": 0.6561797752808989, + "4": 0.052808988764044947, + "5": 0.020786516853932586 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.130859375, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.039640375970576215, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3481814466693911, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18573763792398856, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.46608091540662033, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.36292134831460676, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2095505617977528, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.42752808988764046, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.37279902853673347, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1305403764420158, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49666059502125076, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.30484988452655887, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.20477290223248654, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.49037721324095457, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3053435114503817, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.29770992366412213, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3969465648854962, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.21621621621621623, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.43243243243243246, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.35135135135135137, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6572265625, + "global/total_cost": 0.0, + "num_tokens": 8983781.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8871527777777778, + "simpleqa_verified/avg_subtask_length": 104.57261794634597, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.24853228962818003, + "2": 0.3913894324853229, + "3": 0.3561643835616438, + "4": 0.003913894324853229, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.11284722222222222, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.056429232192414434, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6364477335800185, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.05920444033302498, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.30434782608695654, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9882583170254403, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011741682974559686, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.19010416666666666, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.052083333333333336, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7578125, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5869565217391305, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2391304347826087, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.17391304347826086, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5815972222222222, + "simpleqa_verified/total_cost": 0.0, + "step": 31, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.765625, + "swe_bench/avg_subtask_length": 121.30733944954129, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.004081632653061225, + "2": 0.07551020408163266, + "3": 0.7755102040816326, + "4": 0.09183673469387756, + "5": 0.053061224489795916 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.234375, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.07601572739187418, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.29685452162516385, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5235910878112713, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17955439056356487, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22448979591836735, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7612244897959184, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.014285714285714285, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.47950819672131145, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.38114754098360654, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13934426229508196, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.17960088691796008, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4212860310421286, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3991130820399113, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30985915492957744, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4788732394366197, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2112676056338028, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.23076923076923078, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6153846153846154, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.15384615384615385, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5265625, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.38861491344869137, + "epoch": 0.10457516339869281, + "frac_reward_zero_std": 0.0, + "grad_norm": 1.2786906957626343, + "learning_rate": 9.59581844897906e-07, + "loss": -0.0352, + "reward": 1.1796875, + "reward_std": 0.2672865092754364, + "rewards/MultiTurnTaskReward/mean": 1.1796875, + "rewards/MultiTurnTaskReward/std": 0.49299970269203186, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.000125527381897, + "sampling/importance_sampling_ratio/min": 0.5389323830604553, + "sampling/sampling_logp_difference/max": 1.0382750034332275, + "sampling/sampling_logp_difference/mean": 0.015842726454138756, + "step": 32 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9351851851851852, + "aime/avg_subtask_length": 138.55705263157896, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0049504950495049506, + "2": 0.13985148514851486, + "3": 0.7797029702970297, + "4": 0.06188118811881188, + "5": 0.013613861386138614 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.06481481481481481, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.007157894736842105, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.248, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.019789473684210527, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7322105263157894, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.038366336633663366, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9616336633663366, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3917910447761194, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011194029850746268, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5970149253731343, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3227206946454414, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04775687409551375, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6295224312590448, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.29508196721311475, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.08196721311475409, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6229508196721312, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8181818181818182, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8136574074074074, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 436.0, + "completions/max_terminated_length": 436.0, + "completions/mean_length": 232.078125, + "completions/mean_terminated_length": 232.078125, + "completions/min_length": 94.0, + "completions/min_terminated_length": 94.0, + "epoch": 0.10457516339869281, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8716856060606061, + "global/avg_subtask_length": 126.1201260338716, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07224334600760456, + "2": 0.1906572514937534, + "3": 0.6637696903856599, + "4": 0.053231939163498096, + "5": 0.020097772949483977 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.12831439393939395, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.03820401732965734, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3461992910594722, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18747538400945254, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4663253249310752, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.354155350353069, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21564367191743616, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.43020097772949484, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.37236533957845436, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.13114754098360656, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4964871194379391, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.30950626381724394, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.20265291083271925, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.48784082535003687, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3111111111111111, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2962962962962963, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3925925925925926, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.21621621621621623, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.43243243243243246, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.35135135135135137, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6661931818181818, + "global/total_cost": 0.0, + "num_tokens": 9266127.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8871527777777778, + "simpleqa_verified/avg_subtask_length": 104.57261794634597, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.24853228962818003, + "2": 0.3913894324853229, + "3": 0.3561643835616438, + "4": 0.003913894324853229, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.11284722222222222, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.056429232192414434, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6364477335800185, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.05920444033302498, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.30434782608695654, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9882583170254403, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011741682974559686, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.19010416666666666, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.052083333333333336, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7578125, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5869565217391305, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2391304347826087, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.17391304347826086, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5815972222222222, + "simpleqa_verified/total_cost": 0.0, + "step": 32, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.7767857142857143, + "swe_bench/avg_subtask_length": 122.27003699136868, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0038314176245210726, + "2": 0.07279693486590039, + "3": 0.7854406130268199, + "4": 0.08812260536398467, + "5": 0.04980842911877394 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.22321428571428573, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.07151664611590629, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2965474722564735, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5184956843403206, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.18495684340320592, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2222222222222222, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7605363984674329, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.017241379310344827, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.47692307692307695, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.375, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.14807692307692308, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18464730290456433, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4107883817427386, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4045643153526971, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3055555555555556, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4861111111111111, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20833333333333334, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.23076923076923078, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6153846153846154, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.15384615384615385, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5491071428571429, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.38048681430518627, + "epoch": 0.10784313725490197, + "frac_reward_zero_std": 0.25, + "grad_norm": 0.8693584203720093, + "learning_rate": 9.5633255760077e-07, + "loss": 0.0112, + "reward": 1.44140625, + "reward_std": 0.21673205494880676, + "rewards/MultiTurnTaskReward/mean": 1.44140625, + "rewards/MultiTurnTaskReward/std": 0.27076077461242676, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0003228187561035, + "sampling/importance_sampling_ratio/min": 0.29759037494659424, + "sampling/sampling_logp_difference/max": 1.2120373249053955, + "sampling/sampling_logp_difference/mean": 0.018597092479467392, + "step": 33 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9375, + "aime/avg_subtask_length": 138.62829347385488, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.004761904761904762, + "2": 0.1392857142857143, + "3": 0.7833333333333333, + "4": 0.05952380952380952, + "5": 0.013095238095238096 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.0625, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.006890960680989056, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 14, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2505066882853668, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01905147952979327, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7304418321848399, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.0380952380952381, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9619047619047619, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3911483253588517, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01076555023923445, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5980861244019139, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.33240611961057026, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.0458970792767733, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6216968011126565, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.29508196721311475, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.08196721311475409, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6229508196721312, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8181818181818182, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8203125, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 499.0, + "completions/max_terminated_length": 499.0, + "completions/mean_length": 227.171875, + "completions/mean_terminated_length": 227.171875, + "completions/min_length": 126.0, + "completions/min_terminated_length": 126.0, + "epoch": 0.10784313725490197, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8754595588235294, + "global/avg_subtask_length": 126.4917410290488, + "global/chosen_num_routing_steps_ratio": { + "1": 0.06981627296587926, + "2": 0.1868766404199475, + "3": 0.6719160104986877, + "4": 0.051443569553805774, + "5": 0.01994750656167979 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.12454044117647059, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.036833111828365295, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 14, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.34459844313651034, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19157015378773495, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4638314030757547, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.34540682414698165, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22257217847769029, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.43202099737532806, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3752821670428894, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.13261851015801354, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49209932279909707, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3121468926553672, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2062146892655367, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.481638418079096, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3088235294117647, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3014705882352941, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3897058823529412, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.21052631578947367, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4473684210526316, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.34210526315789475, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6760110294117647, + "global/total_cost": 0.0, + "num_tokens": 9523237.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8871527777777778, + "simpleqa_verified/avg_subtask_length": 104.57261794634597, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.24853228962818003, + "2": 0.3913894324853229, + "3": 0.3561643835616438, + "4": 0.003913894324853229, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.11284722222222222, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.056429232192414434, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6364477335800185, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.05920444033302498, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.30434782608695654, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9882583170254403, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011741682974559686, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.19010416666666666, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.052083333333333336, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7578125, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.5869565217391305, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2391304347826087, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.17391304347826086, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.5815972222222222, + "simpleqa_verified/total_cost": 0.0, + "step": 33, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.7869318181818182, + "swe_bench/avg_subtask_length": 122.85805700988946, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0036101083032490976, + "2": 0.0703971119133574, + "3": 0.7942238267148014, + "4": 0.08303249097472924, + "5": 0.048736462093862815 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.21306818181818182, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.06748109365910412, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2961023851076207, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5223967422920303, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.18150087260034903, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2184115523465704, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7653429602888087, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.016245487364620937, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.48007246376811596, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.37318840579710144, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.14673913043478262, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18518518518518517, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4191033138401559, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.39571150097465885, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3013698630136986, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4931506849315068, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2054794520547945, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2222222222222222, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6296296296296297, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.14814814814814814, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5696022727272727, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4010020736604929, + "epoch": 0.1111111111111111, + "frac_reward_zero_std": 0.5, + "grad_norm": 0.7003167271614075, + "learning_rate": 9.529636049992233e-07, + "loss": 0.0312, + "reward": 1.4765625, + "reward_std": 0.10567784309387207, + "rewards/MultiTurnTaskReward/mean": 1.4765625, + "rewards/MultiTurnTaskReward/std": 0.15188287198543549, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.000075340270996, + "sampling/importance_sampling_ratio/min": 0.3301011323928833, + "sampling/sampling_logp_difference/max": 1.108356237411499, + "sampling/sampling_logp_difference/mean": 0.017808202654123306, + "step": 34 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9396551724137931, + "aime/avg_subtask_length": 138.8081281750684, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0045871559633027525, + "2": 0.14105504587155962, + "3": 0.7821100917431193, + "4": 0.05963302752293578, + "5": 0.01261467889908257 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.0603448275862069, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.006643220007815553, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.25205158264947247, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01875732708089097, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7291910902696366, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.0389908256880734, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9610091743119266, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.38940092165898615, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.010368663594470046, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6002304147465438, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.338255033557047, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04563758389261745, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6161073825503356, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30158730158730157, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.07936507936507936, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6190476190476191, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8181818181818182, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8254310344827587, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 585.0, + "completions/max_terminated_length": 585.0, + "completions/mean_length": 223.3984375, + "completions/mean_terminated_length": 223.3984375, + "completions/min_length": 104.0, + "completions/min_terminated_length": 104.0, + "epoch": 0.1111111111111111, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8790178571428572, + "global/avg_subtask_length": 126.17965726920951, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07110208227526663, + "2": 0.1909598781107161, + "3": 0.6678517013712545, + "4": 0.05078720162519045, + "5": 0.019299136617572373 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.12098214285714286, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.035747189976045694, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3480744426018058, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18647503224617654, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4654505251520177, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.35144743524631794, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21533773489080751, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4332148298628746, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3728813559322034, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12848551120831056, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4986331328594861, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3193392980041294, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.20302821748107364, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.47763248451479695, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3115942028985507, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2971014492753623, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.391304347826087, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.21052631578947367, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4473684210526316, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.34210526315789475, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6848214285714286, + "global/total_cost": 0.0, + "num_tokens": 9732280.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8930921052631579, + "simpleqa_verified/avg_subtask_length": 103.02349869451697, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.24677716390423574, + "2": 0.39410681399631675, + "3": 0.35543278084714547, + "4": 0.003683241252302026, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.1069078947368421, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.05308964316797215, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.639686684073107, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.057441253263707574, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3028720626631854, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.988950276243094, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011049723756906077, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.19315403422982885, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0488997555012225, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7579462102689487, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2358974358974359, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.1641025641025641, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6036184210526315, + "simpleqa_verified/total_cost": 0.0, + "step": 34, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.7869318181818182, + "swe_bench/avg_subtask_length": 122.85805700988946, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0036101083032490976, + "2": 0.0703971119133574, + "3": 0.7942238267148014, + "4": 0.08303249097472924, + "5": 0.048736462093862815 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.21306818181818182, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.06748109365910412, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2961023851076207, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5223967422920303, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.18150087260034903, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2184115523465704, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7653429602888087, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.016245487364620937, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.48007246376811596, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.37318840579710144, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.14673913043478262, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18518518518518517, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4191033138401559, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.39571150097465885, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3013698630136986, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4931506849315068, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2054794520547945, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2222222222222222, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6296296296296297, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.14814814814814814, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5696022727272727, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.361395001411438, + "epoch": 0.11437908496732026, + "frac_reward_zero_std": 0.25, + "grad_norm": 0.9001783728599548, + "learning_rate": 9.494758705426976e-07, + "loss": -0.027, + "reward": 1.22265625, + "reward_std": 0.23535901308059692, + "rewards/MultiTurnTaskReward/mean": 1.22265625, + "rewards/MultiTurnTaskReward/std": 0.4687418043613434, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0002474784851074, + "sampling/importance_sampling_ratio/min": 0.4748454988002777, + "sampling/sampling_logp_difference/max": 1.3712046146392822, + "sampling/sampling_logp_difference/mean": 0.015957936644554138, + "step": 35 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9396551724137931, + "aime/avg_subtask_length": 138.8081281750684, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0045871559633027525, + "2": 0.14105504587155962, + "3": 0.7821100917431193, + "4": 0.05963302752293578, + "5": 0.01261467889908257 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.0603448275862069, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.006643220007815553, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.25205158264947247, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01875732708089097, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7291910902696366, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.0389908256880734, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9610091743119266, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.38940092165898615, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.010368663594470046, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6002304147465438, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.338255033557047, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04563758389261745, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6161073825503356, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30158730158730157, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.07936507936507936, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6190476190476191, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8181818181818182, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8254310344827587, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 713.0, + "completions/max_terminated_length": 713.0, + "completions/mean_length": 244.3984375, + "completions/mean_terminated_length": 244.3984375, + "completions/min_length": 100.0, + "completions/min_terminated_length": 100.0, + "epoch": 0.11437908496732026, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8823784722222222, + "global/avg_subtask_length": 126.17957431586478, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07279881947860305, + "2": 0.191342843089031, + "3": 0.6674864731923266, + "4": 0.04968027545499262, + "5": 0.018691588785046728 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.11762152777777778, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.03469862278662136, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3512788409944554, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19066356644607405, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4580575925594706, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3600590260698475, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22036399409739302, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4195769798327595, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3724137931034483, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.13209549071618037, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4954907161803713, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3201871657754011, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2072192513368984, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.47259358288770054, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30935251798561153, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.302158273381295, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.38848920863309355, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.21052631578947367, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4473684210526316, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.34210526315789475, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.69140625, + "global/total_cost": 0.0, + "num_tokens": 9978475.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.8984375, + "simpleqa_verified/avg_subtask_length": 102.82183908045977, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.24695652173913044, + "2": 0.391304347826087, + "3": 0.3582608695652174, + "4": 0.0034782608695652175, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.1015625, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.05008210180623974, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6403940886699507, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.05747126436781609, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3021346469622332, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9895652173913043, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.010434782608695653, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.18706697459584296, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.050808314087759814, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7621247113163973, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6153846153846154, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.23076923076923078, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.15384615384615385, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6234375, + "simpleqa_verified/total_cost": 0.0, + "step": 35, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.7961956521739131, + "swe_bench/avg_subtask_length": 124.04796030871003, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0034129692832764505, + "2": 0.06996587030716724, + "3": 0.8003412969283277, + "4": 0.08020477815699659, + "5": 0.04607508532423208 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.20380434782608695, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.0639470782800441, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.29713340683572215, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5226019845644984, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1802646085997795, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22013651877133106, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.764505119453925, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.015358361774744027, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4845890410958904, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3732876712328767, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.1421232876712329, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18232044198895028, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4198895027624309, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.39779005524861877, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2972972972972973, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.5, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20270270270270271, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2222222222222222, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6296296296296297, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.14814814814814814, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5815217391304348, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3467192407697439, + "epoch": 0.11764705882352941, + "frac_reward_zero_std": 0.25, + "grad_norm": 0.9950386881828308, + "learning_rate": 9.458702688291071e-07, + "loss": 0.015, + "reward": 1.44140625, + "reward_std": 0.20271094143390656, + "rewards/MultiTurnTaskReward/mean": 1.44140625, + "rewards/MultiTurnTaskReward/std": 0.24799267947673798, + "sampling/importance_sampling_ratio/max": 1.7469537258148193, + "sampling/importance_sampling_ratio/mean": 1.0001161098480225, + "sampling/importance_sampling_ratio/min": 0.20612874627113342, + "sampling/sampling_logp_difference/max": 1.5792542695999146, + "sampling/sampling_logp_difference/mean": 0.017442770302295685, + "step": 36 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9396551724137931, + "aime/avg_subtask_length": 138.8081281750684, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0045871559633027525, + "2": 0.14105504587155962, + "3": 0.7821100917431193, + "4": 0.05963302752293578, + "5": 0.01261467889908257 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.0603448275862069, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.006643220007815553, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.25205158264947247, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01875732708089097, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7291910902696366, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.0389908256880734, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9610091743119266, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.38940092165898615, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.010368663594470046, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6002304147465438, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.338255033557047, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04563758389261745, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6161073825503356, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30158730158730157, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.07936507936507936, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6190476190476191, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8181818181818182, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8254310344827587, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 609.0, + "completions/max_terminated_length": 609.0, + "completions/mean_length": 221.7890625, + "completions/mean_terminated_length": 221.7890625, + "completions/min_length": 104.0, + "completions/min_terminated_length": 104.0, + "epoch": 0.11764705882352941, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8855574324324325, + "global/avg_subtask_length": 125.99255153299845, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07057701478302336, + "2": 0.19074868860276586, + "3": 0.6719122556032427, + "4": 0.04864091559370529, + "5": 0.018121125417262757 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.11444256756756757, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.03360471158842889, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.35510133379525377, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19366014204053353, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.45123852416421273, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3676680972818312, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22556032427277062, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4067715784453982, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3719856336582863, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.13288866085171883, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49512570548999485, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3240800516462234, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21110393802453195, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4648160103292447, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3142857142857143, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.38571428571428573, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.21052631578947367, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4473684210526316, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.34210526315789475, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.6993243243243243, + "global/total_cost": 0.0, + "num_tokens": 10327920.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9032738095238095, + "simpleqa_verified/avg_subtask_length": 101.65003837298542, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.23393739703459637, + "2": 0.38879736408566723, + "3": 0.37397034596375617, + "4": 0.0032948929159802307, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.09672619047619048, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.04681504221028396, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6377590176515733, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.05525709900230238, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3069838833461243, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9901153212520593, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.009884678747940691, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.17849462365591398, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.047311827956989246, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7741935483870968, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6331877729257642, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2183406113537118, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14847161572052403, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6398809523809523, + "simpleqa_verified/total_cost": 0.0, + "step": 36, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8046875, + "swe_bench/avg_subtask_length": 125.429094714809, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.003236245954692557, + "2": 0.06634304207119741, + "3": 0.8090614886731392, + "4": 0.07766990291262135, + "5": 0.043689320388349516 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.1953125, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.06070120355834641, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30036630036630035, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.522239665096808, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17739403453689168, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22006472491909385, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7653721682847896, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.014563106796116505, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4935064935064935, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.37012987012987014, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13636363636363635, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.1826086956521739, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4226086956521739, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3947826086956522, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30666666666666664, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.49333333333333335, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2222222222222222, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6296296296296297, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.14814814814814814, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5989583333333334, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3858554735779762, + "epoch": 0.12091503267973856, + "frac_reward_zero_std": 0.25, + "grad_norm": 1.0718783140182495, + "learning_rate": 9.421477453650117e-07, + "loss": 0.0061, + "reward": 1.15234375, + "reward_std": 0.21720558404922485, + "rewards/MultiTurnTaskReward/mean": 1.15234375, + "rewards/MultiTurnTaskReward/std": 0.484234094619751, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.999967634677887, + "sampling/importance_sampling_ratio/min": 0.3218260705471039, + "sampling/sampling_logp_difference/max": 1.1337440013885498, + "sampling/sampling_logp_difference/mean": 0.020504526793956757, + "step": 37 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9416666666666667, + "aime/avg_subtask_length": 139.25953910086892, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.004424778761061947, + "2": 0.14491150442477876, + "3": 0.7809734513274337, + "4": 0.05752212389380531, + "5": 0.012168141592920354 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.058333333333333334, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.006422364941443143, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.25425009444654323, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.018511522478277295, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7272383830751794, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03761061946902655, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9623893805309734, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.39, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3472041612483745, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.045513654096228866, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6072821846553966, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30158730158730157, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.07936507936507936, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6190476190476191, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8181818181818182, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.83125, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 380.0, + "completions/max_terminated_length": 380.0, + "completions/mean_length": 198.1953125, + "completions/mean_terminated_length": 198.1953125, + "completions/min_length": 99.0, + "completions/min_terminated_length": 99.0, + "epoch": 0.12091503267973856, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8885690789473685, + "global/avg_subtask_length": 125.99410675197845, + "global/chosen_num_routing_steps_ratio": { + "1": 0.07080055529847293, + "2": 0.1925034706154558, + "3": 0.6719111522443313, + "4": 0.047200370198981954, + "5": 0.017584451642757983 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.11143092105263158, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.032665431890890724, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.357635965650783, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18925745074928438, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.45310658359993267, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3715872281351226, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21888014807959277, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4095326237852846, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.36852589641434264, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1294820717131474, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.50199203187251, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.332286432160804, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.20854271356783918, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4591708542713568, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3142857142857143, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.38571428571428573, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.21052631578947367, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4473684210526316, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.34210526315789475, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7060032894736842, + "global/total_cost": 0.0, + "num_tokens": 10532041.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9076704545454546, + "simpleqa_verified/avg_subtask_length": 101.34974656046343, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2300469483568075, + "2": 0.38184663536776214, + "3": 0.38497652582159625, + "4": 0.003129890453834116, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.09232954545454546, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.044170890658942794, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6350470673425054, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.055756698044895005, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.30919623461259954, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9906103286384976, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.009389671361502348, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.17276422764227642, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.046747967479674794, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7804878048780488, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6330645161290323, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21774193548387097, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14919354838709678, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6519886363636364, + "simpleqa_verified/total_cost": 0.0, + "step": 37, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8046875, + "swe_bench/avg_subtask_length": 125.429094714809, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.003236245954692557, + "2": 0.06634304207119741, + "3": 0.8090614886731392, + "4": 0.07766990291262135, + "5": 0.043689320388349516 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.1953125, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.06070120355834641, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30036630036630035, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.522239665096808, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17739403453689168, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.22006472491909385, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7653721682847896, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.014563106796116505, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4935064935064935, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.37012987012987014, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13636363636363635, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.1826086956521739, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4226086956521739, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3947826086956522, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30666666666666664, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.49333333333333335, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2222222222222222, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6296296296296297, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.14814814814814814, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.5989583333333334, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.35869875736534595, + "epoch": 0.12418300653594772, + "frac_reward_zero_std": 0.25, + "grad_norm": 1.0105788707733154, + "learning_rate": 9.383092763176738e-07, + "loss": 0.002, + "reward": 1.4609375, + "reward_std": 0.1624244898557663, + "rewards/MultiTurnTaskReward/mean": 1.4609375, + "rewards/MultiTurnTaskReward/std": 0.194504976272583, + "sampling/importance_sampling_ratio/max": 1.937906265258789, + "sampling/importance_sampling_ratio/mean": 0.9997248649597168, + "sampling/importance_sampling_ratio/min": 0.4596955180168152, + "sampling/sampling_logp_difference/max": 0.7771909236907959, + "sampling/sampling_logp_difference/mean": 0.01680578663945198, + "step": 38 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9416666666666667, + "aime/avg_subtask_length": 139.25953910086892, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.004424778761061947, + "2": 0.14491150442477876, + "3": 0.7809734513274337, + "4": 0.05752212389380531, + "5": 0.012168141592920354 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.058333333333333334, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.006422364941443143, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.25425009444654323, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.018511522478277295, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7272383830751794, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03761061946902655, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9623893805309734, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.39, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3472041612483745, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.045513654096228866, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6072821846553966, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30158730158730157, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.07936507936507936, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6190476190476191, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8181818181818182, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.83125, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 528.0, + "completions/max_terminated_length": 528.0, + "completions/mean_length": 247.25, + "completions/mean_terminated_length": 247.25, + "completions/min_length": 112.0, + "completions/min_terminated_length": 112.0, + "epoch": 0.12418300653594772, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.891426282051282, + "global/avg_subtask_length": 126.26193969030155, + "global/chosen_num_routing_steps_ratio": { + "1": 0.06876404494382023, + "2": 0.18786516853932583, + "3": 0.6786516853932584, + "4": 0.046741573033707864, + "5": 0.017977528089887642 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.10857371794871795, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.03162184189079054, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3568052159739201, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1993480032599837, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4438467807660962, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3653932584269663, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.23685393258426965, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.39775280898876403, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3778957528957529, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.13320463320463322, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4888996138996139, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3258766626360339, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21644498186215236, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4576783555018138, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3194444444444444, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2986111111111111, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3819444444444444, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.475, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.325, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7135416666666666, + "global/total_cost": 0.0, + "num_tokens": 10899977.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9076704545454546, + "simpleqa_verified/avg_subtask_length": 101.34974656046343, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2300469483568075, + "2": 0.38184663536776214, + "3": 0.38497652582159625, + "4": 0.003129890453834116, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.09232954545454546, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.044170890658942794, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6350470673425054, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.055756698044895005, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.30919623461259954, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9906103286384976, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.009389671361502348, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.17276422764227642, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.046747967479674794, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7804878048780488, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6330645161290323, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21774193548387097, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14919354838709678, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6519886363636364, + "simpleqa_verified/total_cost": 0.0, + "step": 38, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8197115384615384, + "swe_bench/avg_subtask_length": 126.26150925486473, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.002932551319648094, + "2": 0.06304985337243402, + "3": 0.8181818181818182, + "4": 0.07331378299120235, + "5": 0.04252199413489736 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.18028846153846154, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.05505457997152349, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3032747982914096, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5206454674893213, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1760797342192691, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.21407624633431085, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7727272727272727, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.013196480938416423, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5102941176470588, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3588235294117647, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13088235294117648, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18053375196232338, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.42229199372056514, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.39717425431711145, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31645569620253167, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4810126582278481, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20253164556962025, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.20689655172413793, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6551724137931034, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.13793103448275862, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6298076923076923, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3916948698461056, + "epoch": 0.12745098039215685, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.6319566369056702, + "learning_rate": 9.343558682590755e-07, + "loss": 0.0048, + "reward": 1.48828125, + "reward_std": 0.06629125773906708, + "rewards/MultiTurnTaskReward/mean": 1.48828125, + "rewards/MultiTurnTaskReward/std": 0.13258251547813416, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9998079538345337, + "sampling/importance_sampling_ratio/min": 0.05624401569366455, + "sampling/sampling_logp_difference/max": 2.8780555725097656, + "sampling/sampling_logp_difference/mean": 0.02098209038376808, + "step": 39 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9435483870967742, + "aime/avg_subtask_length": 139.48705796573094, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.004273504273504274, + "2": 0.14102564102564102, + "3": 0.7863247863247863, + "4": 0.056623931623931624, + "5": 0.011752136752136752 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.056451612903225805, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.006197593875318994, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2548304775792927, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01859278162595698, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7265767407947503, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03632478632478633, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9636752136752137, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3905579399141631, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01072961373390558, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5987124463519313, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.35, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04375, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.60625, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.296875, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.09375, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.609375, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8181818181818182, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8336693548387096, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 366.0, + "completions/max_terminated_length": 366.0, + "completions/mean_length": 196.0390625, + "completions/mean_terminated_length": 196.0390625, + "completions/min_length": 107.0, + "completions/min_terminated_length": 107.0, + "epoch": 0.12745098039215685, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.894140625, + "global/avg_subtask_length": 126.53940066592675, + "global/chosen_num_routing_steps_ratio": { + "1": 0.06771515945827872, + "2": 0.19003931847968544, + "3": 0.6788990825688074, + "4": 0.045871559633027525, + "5": 0.01747487986020096 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.105859375, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.030759473600761058, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.35880767401300145, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19438718883779926, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4468051371491993, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3691568370467453, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.23023154215814767, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.40061162079510704, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.37582005623242737, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12980318650421743, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4943767572633552, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.33078281341965865, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2113007651559741, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4579164214243673, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31724137931034485, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.30344827586206896, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3793103448275862, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.475, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.325, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.708203125, + "global/total_cost": 0.0, + "num_tokens": 11103566.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9116847826086957, + "simpleqa_verified/avg_subtask_length": 102.56554564172959, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.22205663189269748, + "2": 0.38748137108792846, + "3": 0.38748137108792846, + "4": 0.0029806259314456036, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.08831521739130435, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.041866849691146193, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.634866163349348, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.053534660260809885, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.31159917638984214, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9910581222056631, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.00894187779433681, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.1743295019157088, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.044061302681992334, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7816091954022989, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6374045801526718, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2099236641221374, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.15267175572519084, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6277173913043478, + "simpleqa_verified/total_cost": 0.0, + "step": 39, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8197115384615384, + "swe_bench/avg_subtask_length": 126.26150925486473, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.002932551319648094, + "2": 0.06304985337243402, + "3": 0.8181818181818182, + "4": 0.07331378299120235, + "5": 0.04252199413489736 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.18028846153846154, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.05505457997152349, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3032747982914096, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5206454674893213, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1760797342192691, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.21407624633431085, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7727272727272727, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.013196480938416423, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5102941176470588, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3588235294117647, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13088235294117648, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18053375196232338, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.42229199372056514, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.39717425431711145, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31645569620253167, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4810126582278481, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20253164556962025, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.20689655172413793, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6551724137931034, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.13793103448275862, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6298076923076923, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.33121503330767155, + "epoch": 0.13071895424836602, + "frac_reward_zero_std": 0.25, + "grad_norm": 1.0095430612564087, + "learning_rate": 9.302885579019626e-07, + "loss": 0.0204, + "reward": 0.98046875, + "reward_std": 0.22654932737350464, + "rewards/MultiTurnTaskReward/mean": 0.98046875, + "rewards/MultiTurnTaskReward/std": 0.5074342489242554, + "sampling/importance_sampling_ratio/max": 1.9241472482681274, + "sampling/importance_sampling_ratio/mean": 0.9992853403091431, + "sampling/importance_sampling_ratio/min": 0.5151097774505615, + "sampling/sampling_logp_difference/max": 0.6633752584457397, + "sampling/sampling_logp_difference/mean": 0.015884730964899063, + "step": 40 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9453125, + "aime/avg_subtask_length": 139.60366068285816, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.004132231404958678, + "2": 0.13636363636363635, + "3": 0.7923553719008265, + "4": 0.05475206611570248, + "5": 0.012396694214876033 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.0546875, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.00598380851812742, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.25624780007039777, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.018303414290742698, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7254487856388595, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03512396694214876, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9648760330578512, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3921161825726141, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01037344398340249, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5975103734439834, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3545673076923077, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.042067307692307696, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.6033653846153846, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2923076923076923, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.1076923076923077, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8369140625, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 391.0, + "completions/max_terminated_length": 391.0, + "completions/mean_length": 202.5234375, + "completions/mean_terminated_length": 202.5234375, + "completions/min_length": 91.0, + "completions/min_terminated_length": 91.0, + "epoch": 0.13071895424836602, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8967225609756098, + "global/avg_subtask_length": 126.33127794049638, + "global/chosen_num_routing_steps_ratio": { + "1": 0.06629834254143646, + "2": 0.18997025074373142, + "3": 0.6816829579260518, + "4": 0.044623884402889925, + "5": 0.017424564385890354 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.10327743902439024, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.029905965777709265, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.36087559734854324, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1900724525975027, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.44905195005395404, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3727156821079473, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22396940076498087, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.40331491712707185, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3732362312243969, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12608101957214382, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5006827492034592, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.33714285714285713, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.20857142857142857, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4542857142857143, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3150684931506849, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3082191780821918, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3767123287671233, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1951219512195122, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4634146341463415, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.34146341463414637, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7145579268292683, + "global/total_cost": 0.0, + "num_tokens": 11354961.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9153645833333334, + "simpleqa_verified/avg_subtask_length": 101.92592592592592, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.21337126600284495, + "2": 0.3869132290184922, + "3": 0.3968705547652916, + "4": 0.002844950213371266, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.08463541666666667, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.0396361273554256, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6328784925276153, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.05458089668615984, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3125406107862248, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9914651493598862, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.008534850640113799, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.1717902350813743, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.04159132007233273, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.786618444846293, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6405693950177936, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21708185053380782, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.1423487544483986, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6432291666666666, + "simpleqa_verified/total_cost": 0.0, + "step": 40, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8197115384615384, + "swe_bench/avg_subtask_length": 126.26150925486473, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.002932551319648094, + "2": 0.06304985337243402, + "3": 0.8181818181818182, + "4": 0.07331378299120235, + "5": 0.04252199413489736 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.18028846153846154, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.05505457997152349, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3032747982914096, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5206454674893213, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1760797342192691, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.21407624633431085, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7727272727272727, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.013196480938416423, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5102941176470588, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3588235294117647, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13088235294117648, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18053375196232338, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.42229199372056514, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.39717425431711145, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31645569620253167, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4810126582278481, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20253164556962025, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.20689655172413793, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6551724137931034, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.13793103448275862, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6298076923076923, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3617576137185097, + "epoch": 0.13398692810457516, + "frac_reward_zero_std": 0.5, + "grad_norm": 0.8136540055274963, + "learning_rate": 9.261084118279846e-07, + "loss": 0.0086, + "reward": 1.2421875, + "reward_std": 0.10567784309387207, + "rewards/MultiTurnTaskReward/mean": 1.2421875, + "rewards/MultiTurnTaskReward/std": 0.43914902210235596, + "sampling/importance_sampling_ratio/max": 1.6980035305023193, + "sampling/importance_sampling_ratio/mean": 0.9999798536300659, + "sampling/importance_sampling_ratio/min": 0.4872552454471588, + "sampling/sampling_logp_difference/max": 0.7189671993255615, + "sampling/sampling_logp_difference/mean": 0.018415207043290138, + "step": 41 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.946969696969697, + "aime/avg_subtask_length": 139.51923731699011, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.004, + "2": 0.132, + "3": 0.799, + "4": 0.053, + "5": 0.012 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.05303030303030303, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.005788219271365339, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.25672454885938034, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.018726591760299626, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7245488593803201, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.034, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.966, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.39056224899598396, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01104417670682731, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5983935742971888, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3587962962962963, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04282407407407408, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5983796296296297, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2923076923076923, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.1076923076923077, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8380681818181818, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 435.0, + "completions/max_terminated_length": 435.0, + "completions/mean_length": 198.3828125, + "completions/mean_terminated_length": 198.3828125, + "completions/min_length": 91.0, + "completions/min_terminated_length": 91.0, + "epoch": 0.13398692810457516, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.8991815476190477, + "global/avg_subtask_length": 126.85334130662281, + "global/chosen_num_routing_steps_ratio": { + "1": 0.06454282167976831, + "2": 0.18494000827472074, + "3": 0.687215556474969, + "4": 0.04509722796855606, + "5": 0.018204385601985933 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.10081845238095238, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.029002840484377335, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3584990282553446, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19240544177007027, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.44909552997458513, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.36574265618535373, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22838229209764171, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.40587505171700455, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.37372843874391865, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12737726669615215, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49889429455992923, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.33792723263506064, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2089305402425579, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4531422271223815, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3137254901960784, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3006535947712418, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.38562091503267976, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3181818181818182, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7194940476190477, + "global/total_cost": 0.0, + "num_tokens": 11599586.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9153645833333334, + "simpleqa_verified/avg_subtask_length": 101.92592592592592, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.21337126600284495, + "2": 0.3869132290184922, + "3": 0.3968705547652916, + "4": 0.002844950213371266, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.08463541666666667, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.0396361273554256, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6328784925276153, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.05458089668615984, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3125406107862248, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9914651493598862, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.008534850640113799, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.1717902350813743, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.04159132007233273, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.786618444846293, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6405693950177936, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21708185053380782, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.1423487544483986, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6432291666666666, + "simpleqa_verified/total_cost": 0.0, + "step": 41, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8263888888888888, + "swe_bench/avg_subtask_length": 127.3791233619521, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0028011204481792717, + "2": 0.06022408963585434, + "3": 0.8165266106442577, + "4": 0.07563025210084033, + "5": 0.04481792717086835 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.1736111111111111, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.05241753276095797, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3027564392227745, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5187528242205152, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17849073655671036, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.21428571428571427, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.773109243697479, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012605042016806723, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5070224719101124, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.35674157303370785, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13623595505617977, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18385650224215247, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4200298953662182, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3961136023916293, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.313953488372093, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.45348837209302323, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.23255813953488372, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1875, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6875, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.125, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6423611111111112, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.35824072547256947, + "epoch": 0.13725490196078433, + "frac_reward_zero_std": 0.0, + "grad_norm": 1.1480093002319336, + "learning_rate": 9.218165262080022e-07, + "loss": 0.0219, + "reward": 1.1953125, + "reward_std": 0.24642717838287354, + "rewards/MultiTurnTaskReward/mean": 1.1953125, + "rewards/MultiTurnTaskReward/std": 0.46208351850509644, + "sampling/importance_sampling_ratio/max": 1.8180224895477295, + "sampling/importance_sampling_ratio/mean": 0.9998721480369568, + "sampling/importance_sampling_ratio/min": 0.055513739585876465, + "sampling/sampling_logp_difference/max": 2.891124725341797, + "sampling/sampling_logp_difference/mean": 0.017552843317389488, + "step": 42 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9485294117647058, + "aime/avg_subtask_length": 139.7817342565117, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.003875968992248062, + "2": 0.12887596899224807, + "3": 0.8032945736434108, + "4": 0.05232558139534884, + "5": 0.011627906976744186 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.051470588235294115, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.00560501153972964, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.25783053082756346, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01813386086383119, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7240356083086054, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03488372093023256, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9651162790697675, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.39007782101167315, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.010700389105058366, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5992217898832685, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.36201117318435755, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.041340782122905026, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5966480446927375, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2878787878787879, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.10606060606060606, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6060606060606061, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8419117647058824, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 453.0, + "completions/max_terminated_length": 453.0, + "completions/mean_length": 199.4609375, + "completions/mean_terminated_length": 199.4609375, + "completions/min_length": 93.0, + "completions/min_terminated_length": 93.0, + "epoch": 0.13725490196078433, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9015261627906976, + "global/avg_subtask_length": 126.9101238164603, + "global/chosen_num_routing_steps_ratio": { + "1": 0.06328093510681176, + "2": 0.18621523579201935, + "3": 0.6884320838371625, + "4": 0.04433696090286175, + "5": 0.0177347843611447 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.09847383720930232, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.02825928623452294, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.36067006554989073, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18776402039329934, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4515659140568099, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.36960902861749295, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22249093107617895, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4079000403063281, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.37134251290877796, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12392426850258176, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5047332185886403, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.34371643394199786, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.2046186895810956, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.45166487647690656, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3116883116883117, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2987012987012987, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.38961038961038963, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3181818181818182, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7256540697674418, + "global/total_cost": 0.0, + "num_tokens": 11805853.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.91875, + "simpleqa_verified/avg_subtask_length": 102.15565163681285, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.2054421768707483, + "2": 0.3891156462585034, + "3": 0.40272108843537413, + "4": 0.0027210884353741495, + "5": 0.0 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.08125, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.03767757875231625, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6324891908585547, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.053119209388511425, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3143915997529339, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9904761904761905, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.009523809523809525, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.17294520547945205, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.039383561643835614, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7876712328767124, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6476510067114094, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.21140939597315436, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14093959731543623, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6575, + "simpleqa_verified/total_cost": 0.0, + "step": 42, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8263888888888888, + "swe_bench/avg_subtask_length": 127.3791233619521, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0028011204481792717, + "2": 0.06022408963585434, + "3": 0.8165266106442577, + "4": 0.07563025210084033, + "5": 0.04481792717086835 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.1736111111111111, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.05241753276095797, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3027564392227745, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5187528242205152, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17849073655671036, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.21428571428571427, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.773109243697479, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012605042016806723, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5070224719101124, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.35674157303370785, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13623595505617977, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18385650224215247, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4200298953662182, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3961136023916293, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.313953488372093, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.45348837209302323, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.23255813953488372, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1875, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6875, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.125, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6423611111111112, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3662209752947092, + "epoch": 0.14052287581699346, + "frac_reward_zero_std": 0.5, + "grad_norm": 0.8884044289588928, + "learning_rate": 9.174140265146355e-07, + "loss": 0.0167, + "reward": 1.484375, + "reward_std": 0.0883883461356163, + "rewards/MultiTurnTaskReward/mean": 1.484375, + "rewards/MultiTurnTaskReward/std": 0.12450689822435379, + "sampling/importance_sampling_ratio/max": 1.7555235624313354, + "sampling/importance_sampling_ratio/mean": 1.000295877456665, + "sampling/importance_sampling_ratio/min": 0.6171507835388184, + "sampling/sampling_logp_difference/max": 0.5627671480178833, + "sampling/sampling_logp_difference/mean": 0.01633515954017639, + "step": 43 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9485294117647058, + "aime/avg_subtask_length": 139.7817342565117, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.003875968992248062, + "2": 0.12887596899224807, + "3": 0.8032945736434108, + "4": 0.05232558139534884, + "5": 0.011627906976744186 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.051470588235294115, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.00560501153972964, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 5, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.25783053082756346, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01813386086383119, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7240356083086054, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03488372093023256, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9651162790697675, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.39007782101167315, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.010700389105058366, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5992217898832685, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.36201117318435755, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.041340782122905026, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5966480446927375, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2878787878787879, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.10606060606060606, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6060606060606061, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8419117647058824, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 374.0, + "completions/max_terminated_length": 374.0, + "completions/mean_length": 182.953125, + "completions/mean_terminated_length": 182.953125, + "completions/min_length": 84.0, + "completions/min_terminated_length": 84.0, + "epoch": 0.14052287581699346, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9037642045454546, + "global/avg_subtask_length": 126.11120603730599, + "global/chosen_num_routing_steps_ratio": { + "1": 0.062475442043222, + "2": 0.19410609037328094, + "3": 0.6825147347740668, + "4": 0.043222003929273084, + "5": 0.01768172888015717 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.09623579545454546, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.027623522711092127, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 5, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3669372063220846, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18410935497650577, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.44895343870140964, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3854616895874263, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2168958742632613, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3976424361493124, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.36588432523051134, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12154233025984912, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5125733445096395, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3514799154334038, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.20243128964059195, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.44608879492600423, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3161290322580645, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2967741935483871, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3870967741935484, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4888888888888889, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3111111111111111, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7318892045454546, + "global/total_cost": 0.0, + "num_tokens": 12005111.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9247685185185185, + "simpleqa_verified/avg_subtask_length": 101.19921215531795, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.19148936170212766, + "2": 0.3979974968710889, + "3": 0.4067584480600751, + "4": 0.0025031289111389237, + "5": 0.0012515644555694619 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.07523148148148148, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.03432751828925155, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6330894766460327, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.050647158131682614, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.31626336522228476, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9912390488110138, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.008760951188986232, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.17182662538699692, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03869969040247678, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7894736842105263, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6646341463414634, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.19817073170731708, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.13719512195121952, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6828703703703703, + "simpleqa_verified/total_cost": 0.0, + "step": 43, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8263888888888888, + "swe_bench/avg_subtask_length": 127.3791233619521, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0028011204481792717, + "2": 0.06022408963585434, + "3": 0.8165266106442577, + "4": 0.07563025210084033, + "5": 0.04481792717086835 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.1736111111111111, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.05241753276095797, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3027564392227745, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5187528242205152, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17849073655671036, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.21428571428571427, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.773109243697479, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012605042016806723, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5070224719101124, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.35674157303370785, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13623595505617977, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18385650224215247, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4200298953662182, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3961136023916293, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.313953488372093, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.45348837209302323, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.23255813953488372, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1875, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6875, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.125, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6423611111111112, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.31352716870605946, + "epoch": 0.1437908496732026, + "frac_reward_zero_std": 1.0, + "grad_norm": 0.0, + "learning_rate": 9.129020672271281e-07, + "loss": 0.0, + "reward": 1.25, + "reward_std": 0.0, + "rewards/MultiTurnTaskReward/mean": 1.25, + "rewards/MultiTurnTaskReward/std": 0.434714138507843, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.000089168548584, + "sampling/importance_sampling_ratio/min": 0.4440854787826538, + "sampling/sampling_logp_difference/max": 0.8117382526397705, + "sampling/sampling_logp_difference/mean": 0.01578593999147415, + "step": 44 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.95, + "aime/avg_subtask_length": 140.8721636305529, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0037593984962406013, + "2": 0.12593984962406016, + "3": 0.8073308270676691, + "4": 0.05169172932330827, + "5": 0.011278195488721804 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.05, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.005433045701502077, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2588686481303931, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.018536273569830616, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7225950782997763, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03383458646616541, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9661654135338346, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3877358490566038, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011320754716981131, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6009433962264151, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3693304535637149, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.042116630669546434, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5885529157667386, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2835820895522388, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.1044776119402985, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6119402985074627, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8357142857142857, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 841.0, + "completions/max_terminated_length": 841.0, + "completions/mean_length": 209.703125, + "completions/mean_terminated_length": 209.703125, + "completions/min_length": 91.0, + "completions/min_terminated_length": 91.0, + "epoch": 0.1437908496732026, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9059027777777777, + "global/avg_subtask_length": 126.29388888888889, + "global/chosen_num_routing_steps_ratio": { + "1": 0.061709467228823305, + "2": 0.1939440398620161, + "3": 0.6845534687619778, + "4": 0.04254503641241855, + "5": 0.017247987734764277 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.09409722222222222, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.026944444444444444, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3690277777777778, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1802777777777778, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.45069444444444445, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3882713683403603, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21157531621310846, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4001533154465312, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3627450980392157, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.11968954248366014, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5175653594771242, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.35942327497425336, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.19824922760041194, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4423274974253347, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3141025641025641, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2948717948717949, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.391025641025641, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.2, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.4888888888888889, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.3111111111111111, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7336805555555556, + "global/total_cost": 0.0, + "num_tokens": 12263633.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9274553571428571, + "simpleqa_verified/avg_subtask_length": 100.45048439181916, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.1865222623345367, + "2": 0.39590854392298436, + "3": 0.4139590854392298, + "4": 0.0024067388688327317, + "5": 0.0012033694344163659 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.07254464285714286, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.03283100107642627, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6334768568353067, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04951560818083961, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3170075349838536, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9915764139590855, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.00842358604091456, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.17159763313609466, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03994082840236687, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7884615384615384, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6714697406340058, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1873198847262248, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14121037463976946, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6941964285714286, + "simpleqa_verified/total_cost": 0.0, + "step": 44, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8263888888888888, + "swe_bench/avg_subtask_length": 127.3791233619521, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0028011204481792717, + "2": 0.06022408963585434, + "3": 0.8165266106442577, + "4": 0.07563025210084033, + "5": 0.04481792717086835 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.1736111111111111, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.05241753276095797, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3027564392227745, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5187528242205152, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17849073655671036, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.21428571428571427, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.773109243697479, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012605042016806723, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5070224719101124, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.35674157303370785, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.13623595505617977, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18385650224215247, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.4200298953662182, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3961136023916293, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.313953488372093, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.45348837209302323, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.23255813953488372, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1875, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.6875, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.125, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6423611111111112, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3833825998008251, + "epoch": 0.14705882352941177, + "frac_reward_zero_std": 0.5, + "grad_norm": 0.8322101831436157, + "learning_rate": 9.082818315286054e-07, + "loss": -0.0068, + "reward": 1.1328125, + "reward_std": 0.19700348377227783, + "rewards/MultiTurnTaskReward/mean": 1.1328125, + "rewards/MultiTurnTaskReward/std": 0.4839322865009308, + "sampling/importance_sampling_ratio/max": 1.7309693098068237, + "sampling/importance_sampling_ratio/mean": 1.0001575946807861, + "sampling/importance_sampling_ratio/min": 0.2822778820991516, + "sampling/sampling_logp_difference/max": 1.2648632526397705, + "sampling/sampling_logp_difference/mean": 0.017971575260162354, + "step": 45 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.95, + "aime/avg_subtask_length": 140.8721636305529, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0037593984962406013, + "2": 0.12593984962406016, + "3": 0.8073308270676691, + "4": 0.05169172932330827, + "5": 0.011278195488721804 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.05, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.005433045701502077, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2588686481303931, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.018536273569830616, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7225950782997763, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03383458646616541, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9661654135338346, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3877358490566038, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011320754716981131, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.6009433962264151, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3693304535637149, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.042116630669546434, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5885529157667386, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2835820895522388, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.1044776119402985, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6119402985074627, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8357142857142857, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 432.0, + "completions/max_terminated_length": 432.0, + "completions/mean_length": 219.1953125, + "completions/mean_terminated_length": 219.1953125, + "completions/min_length": 107.0, + "completions/min_terminated_length": 107.0, + "epoch": 0.14705882352941177, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9079483695652174, + "global/avg_subtask_length": 126.77148648648648, + "global/chosen_num_routing_steps_ratio": { + "1": 0.06023194912083801, + "2": 0.18967452300785634, + "3": 0.689487467265245, + "4": 0.04264870931537598, + "5": 0.017957351290684626 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.09205163043478261, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.026216216216216216, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3681081081081081, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18932432432432433, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.44256756756756754, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3838383838383838, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2255892255892256, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.39057239057239057, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3678343949044586, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12738853503184713, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5047770700636943, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3561097256857855, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.20199501246882792, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4418952618453865, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3148148148148148, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2962962962962963, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3888888888888889, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1875, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5208333333333334, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2916666666666667, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7381114130434783, + "global/total_cost": 0.0, + "num_tokens": 12590314.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9274553571428571, + "simpleqa_verified/avg_subtask_length": 100.45048439181916, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.1865222623345367, + "2": 0.39590854392298436, + "3": 0.4139590854392298, + "4": 0.0024067388688327317, + "5": 0.0012033694344163659 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.07254464285714286, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.03283100107642627, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6334768568353067, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04951560818083961, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3170075349838536, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9915764139590855, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.00842358604091456, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.17159763313609466, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03994082840236687, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7884615384615384, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6714697406340058, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1873198847262248, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14121037463976946, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6941964285714286, + "simpleqa_verified/total_cost": 0.0, + "step": 45, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8383620689655172, + "swe_bench/avg_subtask_length": 128.75383340240364, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.002570694087403599, + "2": 0.056555269922879174, + "3": 0.8226221079691517, + "4": 0.07326478149100257, + "5": 0.04498714652956298 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.16163793103448276, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.04807293825113966, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30542892664732696, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5184417737256527, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1761292996270203, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2133676092544987, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7750642673521851, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011568123393316195, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5115979381443299, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.36211340206185566, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.12628865979381443, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18989071038251365, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.41120218579234974, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3989071038251366, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31521739130434784, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.44565217391304346, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2391304347826087, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.17142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11428571428571428, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6627155172413793, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3510554004460573, + "epoch": 0.1503267973856209, + "frac_reward_zero_std": 0.25, + "grad_norm": 1.1041010618209839, + "learning_rate": 9.035545309958046e-07, + "loss": 0.042, + "reward": 1.21875, + "reward_std": 0.17239104211330414, + "rewards/MultiTurnTaskReward/mean": 1.21875, + "rewards/MultiTurnTaskReward/std": 0.4513758420944214, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9991657733917236, + "sampling/importance_sampling_ratio/min": 0.28774645924568176, + "sampling/sampling_logp_difference/max": 1.2456755638122559, + "sampling/sampling_logp_difference/mean": 0.01996723935008049, + "step": 46 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9527027027027027, + "aime/avg_subtask_length": 141.1941513415737, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0035460992907801418, + "2": 0.12234042553191489, + "3": 0.8147163120567376, + "4": 0.04875886524822695, + "5": 0.010638297872340425 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.0472972972972973, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0051251130539644255, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2619837202291227, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.018993066023515224, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7190232137473621, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03280141843971631, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9671985815602837, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.398576512455516, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.013345195729537367, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5880782918149466, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.36815415821501013, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04158215010141988, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.59026369168357, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2835820895522388, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.1044776119402985, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6119402985074627, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8420608108108109, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 393.0, + "completions/max_terminated_length": 393.0, + "completions/mean_length": 204.8203125, + "completions/mean_terminated_length": 204.8203125, + "completions/min_length": 102.0, + "completions/min_terminated_length": 102.0, + "epoch": 0.1503267973856209, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.909906914893617, + "global/avg_subtask_length": 127.26159725882974, + "global/chosen_num_routing_steps_ratio": { + "1": 0.058823529411764705, + "2": 0.1867007672634271, + "3": 0.6952868103763244, + "4": 0.04165144318597004, + "5": 0.0175374497625137 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.09009308510638298, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.025566684238270954, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.36676331049024774, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1852925672113864, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.44794412229836583, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3752283522104494, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22031421264157838, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4044574351479722, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3730590062111801, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12538819875776397, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5015527950310559, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3559322033898305, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.19709443099273607, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4469733656174334, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3148148148148148, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.2962962962962963, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3888888888888889, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1875, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5208333333333334, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2916666666666667, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.742686170212766, + "global/total_cost": 0.0, + "num_tokens": 12921427.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9274553571428571, + "simpleqa_verified/avg_subtask_length": 100.45048439181916, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.1865222623345367, + "2": 0.39590854392298436, + "3": 0.4139590854392298, + "4": 0.0024067388688327317, + "5": 0.0012033694344163659 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.07254464285714286, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.03283100107642627, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6334768568353067, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04951560818083961, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3170075349838536, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9915764139590855, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.00842358604091456, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.17159763313609466, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03994082840236687, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7884615384615384, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6714697406340058, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1873198847262248, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14121037463976946, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6941964285714286, + "simpleqa_verified/total_cost": 0.0, + "step": 46, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8383620689655172, + "swe_bench/avg_subtask_length": 128.75383340240364, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.002570694087403599, + "2": 0.056555269922879174, + "3": 0.8226221079691517, + "4": 0.07326478149100257, + "5": 0.04498714652956298 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.16163793103448276, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.04807293825113966, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30542892664732696, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5184417737256527, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1761292996270203, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2133676092544987, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7750642673521851, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011568123393316195, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5115979381443299, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.36211340206185566, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.12628865979381443, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18989071038251365, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.41120218579234974, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3989071038251366, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31521739130434784, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.44565217391304346, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2391304347826087, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.17142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11428571428571428, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6627155172413793, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4028620384633541, + "epoch": 0.15359477124183007, + "frac_reward_zero_std": 0.25, + "grad_norm": 1.1862969398498535, + "learning_rate": 8.987214052813603e-07, + "loss": 0.0369, + "reward": 1.46875, + "reward_std": 0.14987201988697052, + "rewards/MultiTurnTaskReward/mean": 1.46875, + "rewards/MultiTurnTaskReward/std": 0.1746762990951538, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0002093315124512, + "sampling/importance_sampling_ratio/min": 0.37216514348983765, + "sampling/sampling_logp_difference/max": 0.9884176254272461, + "sampling/sampling_logp_difference/mean": 0.017825301736593246, + "step": 47 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9539473684210527, + "aime/avg_subtask_length": 141.88280105479052, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0034482758620689655, + "2": 0.11982758620689656, + "3": 0.8181034482758621, + "4": 0.04827586206896552, + "5": 0.010344827586206896 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.046052631578947366, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.004980955171403457, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.26399062408438323, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.019337825959566363, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7166715499560504, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03189655172413793, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.968103448275862, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4039792387543253, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.012975778546712802, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.583044982698962, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3697148475909538, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04228121927236971, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5880039331366765, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.27941176470588236, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.11764705882352941, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6029411764705882, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8462171052631579, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 399.0, + "completions/max_terminated_length": 399.0, + "completions/mean_length": 200.7109375, + "completions/mean_terminated_length": 200.7109375, + "completions/min_length": 97.0, + "completions/min_terminated_length": 97.0, + "epoch": 0.15359477124183007, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9117838541666666, + "global/avg_subtask_length": 127.51982492276004, + "global/chosen_num_routing_steps_ratio": { + "1": 0.057479471617279544, + "2": 0.1870760442699036, + "3": 0.6972509817922171, + "4": 0.04105676544091396, + "5": 0.017136736879685827 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.08821614583333333, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.024974253347064882, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.368692070030896, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18190010298661174, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.44940782698249226, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3780792574080685, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2152802570510532, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4066404855408783, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3723484848484849, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12310606060606061, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5045454545454545, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3601134215500945, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.194234404536862, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.44565217391304346, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3128834355828221, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3006134969325153, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.38650306748466257, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1875, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5208333333333334, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2916666666666667, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7473958333333334, + "global/total_cost": 0.0, + "num_tokens": 13192398.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9299568965517241, + "simpleqa_verified/avg_subtask_length": 100.74407826982493, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.17960602549246812, + "2": 0.3951332560834299, + "3": 0.42178447276940906, + "4": 0.002317497103128621, + "5": 0.0011587485515643105 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.07004310344827586, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.03141091658084449, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.631307929969104, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04943357363542739, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3192584963954686, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9918887601390498, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.008111239860950173, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.16807909604519775, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.04096045197740113, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7909604519774012, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6730245231607629, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18256130790190736, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.1444141689373297, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7025862068965517, + "simpleqa_verified/total_cost": 0.0, + "step": 47, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8383620689655172, + "swe_bench/avg_subtask_length": 128.75383340240364, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.002570694087403599, + "2": 0.056555269922879174, + "3": 0.8226221079691517, + "4": 0.07326478149100257, + "5": 0.04498714652956298 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.16163793103448276, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.04807293825113966, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30542892664732696, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5184417737256527, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1761292996270203, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2133676092544987, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7750642673521851, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011568123393316195, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5115979381443299, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.36211340206185566, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.12628865979381443, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18989071038251365, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.41120218579234974, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3989071038251366, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31521739130434784, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.44565217391304346, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2391304347826087, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.17142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11428571428571428, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6627155172413793, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.4004264399409294, + "epoch": 0.1568627450980392, + "frac_reward_zero_std": 0.25, + "grad_norm": 1.0060471296310425, + "learning_rate": 8.937837217887272e-07, + "loss": -0.0214, + "reward": 1.4609375, + "reward_std": 0.1671615093946457, + "rewards/MultiTurnTaskReward/mean": 1.4609375, + "rewards/MultiTurnTaskReward/std": 0.194504976272583, + "sampling/importance_sampling_ratio/max": 1.9446614980697632, + "sampling/importance_sampling_ratio/mean": 1.0001481771469116, + "sampling/importance_sampling_ratio/min": 0.584505558013916, + "sampling/sampling_logp_difference/max": 0.6650879383087158, + "sampling/sampling_logp_difference/mean": 0.01783258095383644, + "step": 48 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9551282051282052, + "aime/avg_subtask_length": 141.97492163009403, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.003355704697986577, + "2": 0.11661073825503356, + "3": 0.822986577181208, + "4": 0.04697986577181208, + "5": 0.010067114093959731 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.04487179487179487, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.004844685095468795, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.26560273582217153, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01909375890567113, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7153035052721574, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03104026845637584, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9689597315436241, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4099326599326599, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.012626262626262626, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5774410774410774, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.36892278360343184, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04194470924690181, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5891325071496664, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.27941176470588236, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.11764705882352941, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6029411764705882, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8501602564102564, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 427.0, + "completions/max_terminated_length": 427.0, + "completions/mean_length": 183.3828125, + "completions/mean_terminated_length": 183.3828125, + "completions/min_length": 82.0, + "completions/min_terminated_length": 82.0, + "epoch": 0.1568627450980392, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9135841836734694, + "global/avg_subtask_length": 127.29496221662468, + "global/chosen_num_routing_steps_ratio": { + "1": 0.0568935427574171, + "2": 0.18848167539267016, + "3": 0.6977312390924957, + "4": 0.04013961605584642, + "5": 0.016753926701570682 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.08641581632653061, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.024433249370277077, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3698992443324937, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1787153652392947, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4513853904282116, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3808027923211169, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21047120418848167, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4087260034904014, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3719467061435973, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12102146558105108, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5070318282753515, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3612395929694727, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.19195189639222943, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.44680851063829785, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3128834355828221, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3006134969325153, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.38650306748466257, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1875, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5208333333333334, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2916666666666667, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7525510204081632, + "global/total_cost": 0.0, + "num_tokens": 13429311.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9322916666666666, + "simpleqa_verified/avg_subtask_length": 100.02428146679881, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.17541899441340783, + "2": 0.39888268156424583, + "3": 0.4223463687150838, + "4": 0.0022346368715083797, + "5": 0.0011173184357541898 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.06770833333333333, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.03022794846382557, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6283448959365708, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.050049554013875126, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.321605550049554, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9921787709497206, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.00782122905027933, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.16395663956639567, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.04200542005420054, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7940379403794038, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6692913385826772, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1837270341207349, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14698162729658792, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7125, + "simpleqa_verified/total_cost": 0.0, + "step": 48, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8383620689655172, + "swe_bench/avg_subtask_length": 128.75383340240364, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.002570694087403599, + "2": 0.056555269922879174, + "3": 0.8226221079691517, + "4": 0.07326478149100257, + "5": 0.04498714652956298 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.16163793103448276, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.04807293825113966, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30542892664732696, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5184417737256527, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1761292996270203, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2133676092544987, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7750642673521851, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011568123393316195, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5115979381443299, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.36211340206185566, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.12628865979381443, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18989071038251365, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.41120218579234974, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3989071038251366, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31521739130434784, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.44565217391304346, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2391304347826087, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.17142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11428571428571428, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6627155172413793, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.37169688753783703, + "epoch": 0.16013071895424835, + "frac_reward_zero_std": 1.0, + "grad_norm": 0.0, + "learning_rate": 8.887427753398247e-07, + "loss": 0.0, + "reward": 1.5, + "reward_std": 0.0, + "rewards/MultiTurnTaskReward/mean": 1.5, + "rewards/MultiTurnTaskReward/std": 0.0, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0001047849655151, + "sampling/importance_sampling_ratio/min": 0.25168076157569885, + "sampling/sampling_logp_difference/max": 1.379593849182129, + "sampling/sampling_logp_difference/mean": 0.018064120784401894, + "step": 49 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9551282051282052, + "aime/avg_subtask_length": 141.97492163009403, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.003355704697986577, + "2": 0.11661073825503356, + "3": 0.822986577181208, + "4": 0.04697986577181208, + "5": 0.010067114093959731 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.04487179487179487, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.004844685095468795, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.26560273582217153, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01909375890567113, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7153035052721574, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03104026845637584, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9689597315436241, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4099326599326599, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.012626262626262626, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5774410774410774, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.36892278360343184, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.04194470924690181, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5891325071496664, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.27941176470588236, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.11764705882352941, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6029411764705882, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8501602564102564, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 392.0, + "completions/max_terminated_length": 392.0, + "completions/mean_length": 200.328125, + "completions/mean_terminated_length": 200.328125, + "completions/min_length": 87.0, + "completions/min_terminated_length": 87.0, + "epoch": 0.16013071895424835, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9153125, + "global/avg_subtask_length": 127.09995076317085, + "global/chosen_num_routing_steps_ratio": { + "1": 0.05565039262546944, + "2": 0.18846022533287812, + "3": 0.699214749061113, + "4": 0.0399453738477296, + "5": 0.016729259132809832 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.0846875, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.023879862136878385, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3709995076317085, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18168389955686853, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.44731659281142294, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3885285080232161, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2116763400477979, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.399795151928986, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3665943600867679, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.12762111352133043, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5057845263919016, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.36224028906955735, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1919602529358627, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.44579945799457993, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3072289156626506, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3132530120481928, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3795180722891566, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1836734693877551, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5306122448979592, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2857142857142857, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7515625, + "global/total_cost": 0.0, + "num_tokens": 13743209.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9344758064516129, + "simpleqa_verified/avg_subtask_length": 99.66460513796385, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.16936353829557713, + "2": 0.39805825242718446, + "3": 0.42934196332254587, + "4": 0.002157497303128371, + "5": 0.0010787486515641855 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.0655241935483871, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.029019980970504282, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6270218839200761, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.049476688867745006, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3235014272121789, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9924487594390508, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007551240560949299, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15844155844155844, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.04285714285714286, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7987012987012987, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6783042394014963, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1770573566084788, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14463840399002495, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7217741935483871, + "simpleqa_verified/total_cost": 0.0, + "step": 49, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.84375, + "swe_bench/avg_subtask_length": 129.27775567051333, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0024691358024691358, + "2": 0.05432098765432099, + "3": 0.825925925925926, + "4": 0.0728395061728395, + "5": 0.044444444444444446 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.15625, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.046159968165539196, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3040191006764823, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5192996418623159, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17668125746120175, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2234567901234568, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7654320987654321, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011111111111111112, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.5012376237623762, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3774752475247525, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.12128712871287128, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18717277486910994, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.40575916230366493, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.40706806282722513, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30526315789473685, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4631578947368421, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.23157894736842105, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7222222222222222, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.1111111111111111, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6541666666666667, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.37723567709326744, + "epoch": 0.16339869281045752, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.6241621971130371, + "learning_rate": 8.83599887835493e-07, + "loss": 0.0008, + "reward": 1.1015625, + "reward_std": 0.12474773079156876, + "rewards/MultiTurnTaskReward/mean": 1.1015625, + "rewards/MultiTurnTaskReward/std": 0.4915000796318054, + "sampling/importance_sampling_ratio/max": 1.7525767087936401, + "sampling/importance_sampling_ratio/mean": 1.0001888275146484, + "sampling/importance_sampling_ratio/min": 0.25482070446014404, + "sampling/sampling_logp_difference/max": 1.3671951293945312, + "sampling/sampling_logp_difference/mean": 0.01898990012705326, + "step": 50 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.95625, + "aime/avg_subtask_length": 142.0249653259362, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0032679738562091504, + "2": 0.11437908496732026, + "3": 0.8259803921568627, + "4": 0.04656862745098039, + "5": 0.00980392156862745 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.04375, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.004715672676837725, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.26712898751733705, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.018585298196948683, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7142857142857143, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.03022875816993464, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9697712418300654, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4131147540983607, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.012295081967213115, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5745901639344262, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3712962962962963, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.040740740740740744, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5879629629629629, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2753623188405797, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.11594202898550725, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6086956521739131, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.85390625, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 319.0, + "completions/max_terminated_length": 319.0, + "completions/mean_length": 199.9765625, + "completions/mean_terminated_length": 199.9765625, + "completions/min_length": 103.0, + "completions/min_terminated_length": 103.0, + "epoch": 0.16339869281045752, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9169730392156863, + "global/avg_subtask_length": 127.36104769914694, + "global/chosen_num_routing_steps_ratio": { + "1": 0.054460407617774806, + "2": 0.18476445038422987, + "3": 0.7029736050785166, + "4": 0.0410958904109589, + "5": 0.01670564650851988 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.08302696078431372, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.023308903039769315, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3700588730025231, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18286675477592215, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4470743722215547, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3855663214166388, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21216171065820247, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4022719679251587, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.36784452296819786, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.13250883392226148, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49964664310954066, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3614404918752745, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1888449714536671, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.44971453667105843, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3063583815028902, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3179190751445087, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.37572254335260113, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.54, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.28, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7564338235294118, + "global/total_cost": 0.0, + "num_tokens": 14059622.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9344758064516129, + "simpleqa_verified/avg_subtask_length": 99.66460513796385, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.16936353829557713, + "2": 0.39805825242718446, + "3": 0.42934196332254587, + "4": 0.002157497303128371, + "5": 0.0010787486515641855 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.0655241935483871, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.029019980970504282, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6270218839200761, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.049476688867745006, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3235014272121789, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9924487594390508, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007551240560949299, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15844155844155844, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.04285714285714286, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7987012987012987, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6783042394014963, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1770573566084788, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14463840399002495, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7217741935483871, + "simpleqa_verified/total_cost": 0.0, + "step": 50, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8487903225806451, + "swe_bench/avg_subtask_length": 129.407874617737, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0023752969121140144, + "2": 0.052256532066508314, + "3": 0.8254156769596199, + "4": 0.07600950118764846, + "5": 0.043942992874109264 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.15120967741935484, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.04434250764525994, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3054281345565749, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5164373088685015, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17813455657492355, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2339667458432304, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7541567695961995, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011876484560570071, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.49404761904761907, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3892857142857143, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.11666666666666667, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.1884422110552764, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3957286432160804, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4158291457286432, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3069306930693069, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46534653465346537, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.22772277227722773, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16216216216216217, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7297297297297297, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.10810810810810811, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6653225806451613, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.36352184414863586, + "epoch": 0.16666666666666666, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.6561912894248962, + "learning_rate": 8.783564079088476e-07, + "loss": 0.0104, + "reward": 1.4921875, + "reward_std": 0.04419417306780815, + "rewards/MultiTurnTaskReward/mean": 1.4921875, + "rewards/MultiTurnTaskReward/std": 0.0883883461356163, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9994806051254272, + "sampling/importance_sampling_ratio/min": 0.16076865792274475, + "sampling/sampling_logp_difference/max": 1.8277888298034668, + "sampling/sampling_logp_difference/mean": 0.019072173163294792, + "step": 51 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9573170731707317, + "aime/avg_subtask_length": 143.15332612222824, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0031847133757961785, + "2": 0.11464968152866242, + "3": 0.8264331210191083, + "4": 0.04617834394904458, + "5": 0.009554140127388535 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.042682926829268296, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.004597079502433748, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.26717144402379667, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01811790156841536, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.714710654407788, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.02945859872611465, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9705414012738853, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.41134185303514376, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011980830670926517, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5766773162939297, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.37454873646209386, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.039711191335740074, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5857400722021661, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2714285714285714, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.11428571428571428, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6142857142857143, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8574695121951219, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 301.0, + "completions/max_terminated_length": 301.0, + "completions/mean_length": 177.5078125, + "completions/mean_terminated_length": 177.5078125, + "completions/min_length": 86.0, + "completions/min_terminated_length": 86.0, + "epoch": 0.16666666666666666, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9185697115384616, + "global/avg_subtask_length": 128.22962179938924, + "global/chosen_num_routing_steps_ratio": { + "1": 0.05332024860974812, + "2": 0.1822047759241086, + "3": 0.7069021916912005, + "4": 0.04121687929342493, + "5": 0.016355904481517827 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.08143028846153846, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.022785999530185577, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3682170542635659, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18404980032887008, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.44773314540756404, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.38141969250899577, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21360811252862283, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.4049721949623814, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.36800276434001383, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.13476157567380787, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4972356599861783, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.36029097133076593, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.188275566966196, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4514334617030381, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30113636363636365, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.32386363636363635, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.375, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.54, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.28, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7605168269230769, + "global/total_cost": 0.0, + "num_tokens": 14294983.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9344758064516129, + "simpleqa_verified/avg_subtask_length": 99.66460513796385, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.16936353829557713, + "2": 0.39805825242718446, + "3": 0.42934196332254587, + "4": 0.002157497303128371, + "5": 0.0010787486515641855 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.0655241935483871, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.029019980970504282, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6270218839200761, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.049476688867745006, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3235014272121789, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9924487594390508, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007551240560949299, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15844155844155844, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.04285714285714286, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7987012987012987, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6783042394014963, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1770573566084788, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14463840399002495, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7217741935483871, + "simpleqa_verified/total_cost": 0.0, + "step": 51, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.853515625, + "swe_bench/avg_subtask_length": 130.01879145173177, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.002288329519450801, + "2": 0.05034324942791762, + "3": 0.8295194508009154, + "4": 0.07551487414187644, + "5": 0.04233409610983982 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.146484375, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.04274134119380987, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30545320560058953, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5143699336772292, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.18017686072218128, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2391304347826087, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7471395881006865, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.013729977116704805, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4908256880733945, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3922018348623853, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.11697247706422019, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18719806763285024, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.392512077294686, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.42028985507246375, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30097087378640774, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.47572815533980584, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.22330097087378642, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16216216216216217, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7297297297297297, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.10810810810810811, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.673828125, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.34544069692492485, + "epoch": 0.16993464052287582, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.7626632452011108, + "learning_rate": 8.73013710571623e-07, + "loss": 0.011, + "reward": 1.484375, + "reward_std": 0.06148367002606392, + "rewards/MultiTurnTaskReward/mean": 1.484375, + "rewards/MultiTurnTaskReward/std": 0.12450689822435379, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.999843418598175, + "sampling/importance_sampling_ratio/min": 0.3210769295692444, + "sampling/sampling_logp_difference/max": 1.1360745429992676, + "sampling/sampling_logp_difference/mean": 0.016720375046133995, + "step": 52 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9573170731707317, + "aime/avg_subtask_length": 143.15332612222824, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0031847133757961785, + "2": 0.11464968152866242, + "3": 0.8264331210191083, + "4": 0.04617834394904458, + "5": 0.009554140127388535 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.042682926829268296, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.004597079502433748, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.26717144402379667, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01811790156841536, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.714710654407788, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.02945859872611465, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9705414012738853, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.41134185303514376, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011980830670926517, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5766773162939297, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.37454873646209386, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.039711191335740074, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5857400722021661, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2714285714285714, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.11428571428571428, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6142857142857143, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8574695121951219, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 365.0, + "completions/max_terminated_length": 365.0, + "completions/mean_length": 181.9765625, + "completions/mean_terminated_length": 181.9765625, + "completions/min_length": 82.0, + "completions/min_terminated_length": 82.0, + "epoch": 0.16993464052287582, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9201061320754716, + "global/avg_subtask_length": 127.56693548387096, + "global/chosen_num_routing_steps_ratio": { + "1": 0.05222685036847164, + "2": 0.1867991028516501, + "3": 0.7045818647869273, + "4": 0.04037167574495354, + "5": 0.016020506247997435 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.07989386792452831, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.022350230414746545, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.37327188940092165, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18099078341013825, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4457373271889401, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.393784043575777, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.20954822172380647, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.39666773470041655, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.36375929682217717, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.13218390804597702, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5040567951318459, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.36757894736842106, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18610526315789475, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4463157894736842, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30113636363636365, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.32386363636363635, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.375, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.54, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.28, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7647405660377359, + "global/total_cost": 0.0, + "num_tokens": 14527556.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9384469696969697, + "simpleqa_verified/avg_subtask_length": 99.21913580246914, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.1584258324924319, + "2": 0.3985872855701312, + "3": 0.4399596367305752, + "4": 0.0020181634712411706, + "5": 0.0010090817356205853 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.061553030303030304, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.026895943562610228, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6274250440917107, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.047619047619047616, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.32495590828924165, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9919273461150353, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0010090817356205853, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.007063572149344097, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15947242206235013, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0407673860911271, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.7997601918465228, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6902050113895216, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1662870159453303, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14350797266514806, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7376893939393939, + "simpleqa_verified/total_cost": 0.0, + "step": 52, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.853515625, + "swe_bench/avg_subtask_length": 130.01879145173177, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.002288329519450801, + "2": 0.05034324942791762, + "3": 0.8295194508009154, + "4": 0.07551487414187644, + "5": 0.04233409610983982 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.146484375, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.04274134119380987, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30545320560058953, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5143699336772292, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.18017686072218128, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2391304347826087, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7471395881006865, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.013729977116704805, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4908256880733945, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.3922018348623853, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.11697247706422019, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.18719806763285024, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.392512077294686, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.42028985507246375, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.30097087378640774, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.47572815533980584, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.22330097087378642, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16216216216216217, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7297297297297297, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.10810810810810811, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.673828125, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3015756197273731, + "epoch": 0.17320261437908496, + "frac_reward_zero_std": 0.25, + "grad_norm": 1.282296061515808, + "learning_rate": 8.675731968536002e-07, + "loss": -0.0028, + "reward": 1.2421875, + "reward_std": 0.1671615093946457, + "rewards/MultiTurnTaskReward/mean": 1.2421875, + "rewards/MultiTurnTaskReward/std": 0.43914902210235596, + "sampling/importance_sampling_ratio/max": 1.814653992652893, + "sampling/importance_sampling_ratio/mean": 0.999890923500061, + "sampling/importance_sampling_ratio/min": 0.23333138227462769, + "sampling/sampling_logp_difference/max": 1.4552955627441406, + "sampling/sampling_logp_difference/mean": 0.01682468131184578, + "step": 53 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9573170731707317, + "aime/avg_subtask_length": 143.15332612222824, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0031847133757961785, + "2": 0.11464968152866242, + "3": 0.8264331210191083, + "4": 0.04617834394904458, + "5": 0.009554140127388535 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.042682926829268296, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.004597079502433748, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.26717144402379667, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01811790156841536, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.714710654407788, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.02945859872611465, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9705414012738853, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.41134185303514376, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011980830670926517, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5766773162939297, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.37454873646209386, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.039711191335740074, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5857400722021661, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2714285714285714, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.11428571428571428, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6142857142857143, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8574695121951219, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 290.0, + "completions/max_terminated_length": 290.0, + "completions/mean_length": 180.4921875, + "completions/mean_terminated_length": 180.4921875, + "completions/min_length": 96.0, + "completions/min_terminated_length": 96.0, + "epoch": 0.17320261437908496, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9212962962962963, + "global/avg_subtask_length": 127.6270440960866, + "global/chosen_num_routing_steps_ratio": { + "1": 0.05119346733668342, + "2": 0.1875, + "3": 0.7032035175879398, + "4": 0.0414572864321608, + "5": 0.01664572864321608 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.0787037037037037, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.02187887673395737, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3739709033494981, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18484267508740274, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4411864215630991, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.39824120603015073, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2129396984924623, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3888190954773869, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3598146309169149, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.13704071499503476, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5031446540880503, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3688118811881188, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18976897689768976, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4414191419141914, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3081081081081081, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.31891891891891894, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.372972972972973, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18867924528301888, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5283018867924528, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2830188679245283, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7598379629629629, + "global/total_cost": 0.0, + "num_tokens": 14775939.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9393382352941176, + "simpleqa_verified/avg_subtask_length": 99.6071580741372, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.15362035225048923, + "2": 0.40019569471624267, + "3": 0.4432485322896282, + "4": 0.0019569471624266144, + "5": 0.0009784735812133072 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.06066176470588235, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.025990626331487004, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6276097145291862, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.046868342564976566, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.32552194290583725, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9921722113502935, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0009784735812133072, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.00684931506849315, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15953757225433526, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03930635838150289, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8011560693641618, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6951754385964912, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.16447368421052633, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14035087719298245, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7159926470588235, + "simpleqa_verified/total_cost": 0.0, + "step": 53, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8579545454545454, + "swe_bench/avg_subtask_length": 130.5846917080085, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.002207505518763797, + "2": 0.04856512141280353, + "3": 0.82560706401766, + "4": 0.07947019867549669, + "5": 0.04415011037527594 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.14204545454545456, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.041105598866052445, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30297661233167966, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5180722891566265, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17895109851169383, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.23951434878587197, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7472406181015453, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.013245033112582781, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.48008849557522126, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.40376106194690264, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.1161504424778761, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.1883720930232558, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.39651162790697675, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41511627906976745, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3125, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.45535714285714285, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.23214285714285715, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.175, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.125, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6837121212121212, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.347843524068594, + "epoch": 0.17647058823529413, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.6417975425720215, + "learning_rate": 8.620362934352108e-07, + "loss": 0.0088, + "reward": 0.99609375, + "reward_std": 0.022097086533904076, + "rewards/MultiTurnTaskReward/mean": 0.99609375, + "rewards/MultiTurnTaskReward/std": 0.5077978372573853, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9999184012413025, + "sampling/importance_sampling_ratio/min": 0.29778456687927246, + "sampling/sampling_logp_difference/max": 1.2113850116729736, + "sampling/sampling_logp_difference/mean": 0.018180711194872856, + "step": 54 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9583333333333334, + "aime/avg_subtask_length": 143.05060622034793, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.003105590062111801, + "2": 0.1125776397515528, + "3": 0.8291925465838509, + "4": 0.04580745341614907, + "5": 0.009316770186335404 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.041666666666666664, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0044807590933052185, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.26884554559831314, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.017923036373220874, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.713231418028466, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.02872670807453416, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9712732919254659, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4135514018691589, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011682242990654205, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5747663551401869, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3775241439859526, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03950834064969271, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5829675153643546, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.28169014084507044, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.11267605633802817, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6056338028169014, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8608630952380952, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 288.0, + "completions/max_terminated_length": 288.0, + "completions/mean_length": 174.2265625, + "completions/mean_terminated_length": 174.2265625, + "completions/min_length": 96.0, + "completions/min_terminated_length": 96.0, + "epoch": 0.17647058823529413, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9227272727272727, + "global/avg_subtask_length": 127.4768380320619, + "global/chosen_num_routing_steps_ratio": { + "1": 0.050492610837438424, + "2": 0.18780788177339902, + "3": 0.7044334975369458, + "4": 0.040948275862068964, + "5": 0.016317733990147784 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.07727272727272727, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.02144831398562742, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.375787728026534, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18153676064123825, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.44267551133222777, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4002463054187192, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.20874384236453203, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3910098522167488, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.35992217898832685, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.13424124513618677, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5058365758754864, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.37227162489894905, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18714632174616005, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.44058205335489087, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3118279569892473, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3172043010752688, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3709677419354839, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18867924528301888, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5283018867924528, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2830188679245283, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7556818181818182, + "global/total_cost": 0.0, + "num_tokens": 14975008.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9410714285714286, + "simpleqa_verified/avg_subtask_length": 99.54055166735282, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.14990512333965844, + "2": 0.39943074003795065, + "3": 0.4478178368121442, + "4": 0.0018975332068311196, + "5": 0.0009487666034155598 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.05892857142857143, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.025113215314944422, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6274186908192672, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04610951008645533, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3264717990942775, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9924098671726755, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0009487666034155598, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.006641366223908918, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.16183035714285715, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03794642857142857, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8002232142857143, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.6926315789473684, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.16210526315789472, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14526315789473684, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.6973214285714285, + "simpleqa_verified/total_cost": 0.0, + "step": 54, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8579545454545454, + "swe_bench/avg_subtask_length": 130.5846917080085, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.002207505518763797, + "2": 0.04856512141280353, + "3": 0.82560706401766, + "4": 0.07947019867549669, + "5": 0.04415011037527594 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.14204545454545456, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.041105598866052445, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30297661233167966, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5180722891566265, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17895109851169383, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.23951434878587197, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7472406181015453, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.013245033112582781, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.48008849557522126, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.40376106194690264, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.1161504424778761, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.1883720930232558, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.39651162790697675, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41511627906976745, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3125, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.45535714285714285, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.23214285714285715, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.175, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.125, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6837121212121212, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3020884720608592, + "epoch": 0.17973856209150327, + "frac_reward_zero_std": 0.5, + "grad_norm": 2.112973690032959, + "learning_rate": 8.564044522734146e-07, + "loss": -0.0086, + "reward": 1.1875, + "reward_std": 0.1792159378528595, + "rewards/MultiTurnTaskReward/mean": 1.1875, + "rewards/MultiTurnTaskReward/std": 0.4653336703777313, + "sampling/importance_sampling_ratio/max": 1.758626103401184, + "sampling/importance_sampling_ratio/mean": 0.9996308088302612, + "sampling/importance_sampling_ratio/min": 0.2877035439014435, + "sampling/sampling_logp_difference/max": 1.2458248138427734, + "sampling/sampling_logp_difference/mean": 0.015416478738188744, + "step": 55 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9583333333333334, + "aime/avg_subtask_length": 143.05060622034793, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.003105590062111801, + "2": 0.1125776397515528, + "3": 0.8291925465838509, + "4": 0.04580745341614907, + "5": 0.009316770186335404 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.041666666666666664, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0044807590933052185, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.26884554559831314, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.017923036373220874, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.713231418028466, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.02872670807453416, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9712732919254659, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4135514018691589, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011682242990654205, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5747663551401869, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3775241439859526, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03950834064969271, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5829675153643546, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.28169014084507044, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.11267605633802817, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6056338028169014, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8608630952380952, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 325.0, + "completions/max_terminated_length": 325.0, + "completions/mean_length": 166.921875, + "completions/mean_terminated_length": 166.921875, + "completions/min_length": 80.0, + "completions/min_terminated_length": 80.0, + "epoch": 0.17973856209150327, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9241071428571429, + "global/avg_subtask_length": 126.92000434169108, + "global/chosen_num_routing_steps_ratio": { + "1": 0.049818840579710144, + "2": 0.19082125603864733, + "3": 0.7032004830917874, + "4": 0.04015700483091787, + "5": 0.0160024154589372 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.07589285714285714, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.021057201780093345, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.380332139368284, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1783349614674916, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.44133289916422447, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.41183574879227053, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.20471014492753623, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3834541062801932, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3549412138544646, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1315538608198284, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.513504925325707, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3797216699801193, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18449304174950298, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4357852882703777, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3118279569892473, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3172043010752688, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3709677419354839, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18867924528301888, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5283018867924528, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2830188679245283, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7600446428571429, + "global/total_cost": 0.0, + "num_tokens": 15172758.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9442567567567568, + "simpleqa_verified/avg_subtask_length": 99.37235271467077, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.14221824686940965, + "2": 0.3962432915921288, + "3": 0.4588550983899821, + "4": 0.0017889087656529517, + "5": 0.0008944543828264759 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.05574324324324324, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.023488640739314594, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.627262225644975, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04351174432036966, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3292260300346554, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9928443649373881, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0008944543828264759, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.006261180679785331, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15849843587069865, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.035453597497393116, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8060479666319083, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7034883720930233, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1511627906976744, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14534883720930233, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7136824324324325, + "simpleqa_verified/total_cost": 0.0, + "step": 55, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8579545454545454, + "swe_bench/avg_subtask_length": 130.5846917080085, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.002207505518763797, + "2": 0.04856512141280353, + "3": 0.82560706401766, + "4": 0.07947019867549669, + "5": 0.04415011037527594 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.14204545454545456, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.041105598866052445, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30297661233167966, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5180722891566265, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17895109851169383, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.23951434878587197, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7472406181015453, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.013245033112582781, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.48008849557522126, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.40376106194690264, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.1161504424778761, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.1883720930232558, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.39651162790697675, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41511627906976745, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3125, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.45535714285714285, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.23214285714285715, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.175, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.125, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6837121212121212, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3199602421373129, + "epoch": 0.1830065359477124, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.5970574617385864, + "learning_rate": 8.506791502209496e-07, + "loss": 0.004, + "reward": 1.46875, + "reward_std": 0.08400268852710724, + "rewards/MultiTurnTaskReward/mean": 1.46875, + "rewards/MultiTurnTaskReward/std": 0.1746762990951538, + "sampling/importance_sampling_ratio/max": 1.69774329662323, + "sampling/importance_sampling_ratio/mean": 1.0002363920211792, + "sampling/importance_sampling_ratio/min": 0.29998528957366943, + "sampling/sampling_logp_difference/max": 1.2040218114852905, + "sampling/sampling_logp_difference/mean": 0.016690116375684738, + "step": 56 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9593023255813954, + "aime/avg_subtask_length": 143.35142783637767, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0030303030303030303, + "2": 0.11212121212121212, + "3": 0.831060606060606, + "4": 0.0446969696969697, + "5": 0.00909090909090909 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.040697674418604654, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.004373552868536146, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.26935940313866735, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01775147928994083, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7128891175713918, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.02803030303030303, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9719696969696969, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.41413373860182373, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011398176291793313, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.574468085106383, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3792808219178082, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.039383561643835614, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5813356164383562, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.28169014084507044, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.11267605633802817, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6056338028169014, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8640988372093024, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 326.0, + "completions/max_terminated_length": 326.0, + "completions/mean_length": 176.8515625, + "completions/mean_terminated_length": 176.8515625, + "completions/min_length": 86.0, + "completions/min_terminated_length": 86.0, + "epoch": 0.1830065359477124, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9254385964912281, + "global/avg_subtask_length": 127.40308346624136, + "global/chosen_num_routing_steps_ratio": { + "1": 0.04887440758293839, + "2": 0.18809241706161137, + "3": 0.7070497630331753, + "4": 0.04028436018957346, + "5": 0.01569905213270142 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.07456140350877193, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.020627325890483786, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.37884104199893676, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18054226475279106, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4406166932482722, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.40580568720379145, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.20853080568720378, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.38566350710900477, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3562753036437247, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.13453752725007787, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5091871691061974, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3804347826086957, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18361801242236025, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4359472049689441, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31216931216931215, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.32275132275132273, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.36507936507936506, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18867924528301888, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5283018867924528, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2830188679245283, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7628837719298246, + "global/total_cost": 0.0, + "num_tokens": 15412163.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9442567567567568, + "simpleqa_verified/avg_subtask_length": 99.37235271467077, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.14221824686940965, + "2": 0.3962432915921288, + "3": 0.4588550983899821, + "4": 0.0017889087656529517, + "5": 0.0008944543828264759 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.05574324324324324, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.023488640739314594, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.627262225644975, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04351174432036966, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3292260300346554, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9928443649373881, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0008944543828264759, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.006261180679785331, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15849843587069865, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.035453597497393116, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8060479666319083, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7034883720930233, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1511627906976744, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14534883720930233, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7136824324324325, + "simpleqa_verified/total_cost": 0.0, + "step": 56, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8621323529411765, + "swe_bench/avg_subtask_length": 131.10201985621362, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0021321961620469083, + "2": 0.046908315565031986, + "3": 0.8283582089552238, + "4": 0.07995735607675906, + "5": 0.042643923240938165 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.13786764705882354, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.03971242725094146, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30366312906538856, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5190003423485108, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17733652858610066, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.23773987206823027, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7494669509594882, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.01279317697228145, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4775641025641026, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4091880341880342, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.11324786324786325, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.19506726457399104, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.39125560538116594, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.413677130044843, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3130434782608696, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4608695652173913, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.22608695652173913, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.175, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.125, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6884191176470589, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.36140489391982555, + "epoch": 0.18627450980392157, + "frac_reward_zero_std": 0.5, + "grad_norm": 1.247727394104004, + "learning_rate": 8.448618886390521e-07, + "loss": 0.0183, + "reward": 1.3671875, + "reward_std": 0.21519285440444946, + "rewards/MultiTurnTaskReward/mean": 1.3671875, + "rewards/MultiTurnTaskReward/std": 0.3407054841518402, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0004422664642334, + "sampling/importance_sampling_ratio/min": 0.35438886284828186, + "sampling/sampling_logp_difference/max": 1.224229335784912, + "sampling/sampling_logp_difference/mean": 0.018051860854029655, + "step": 57 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9593023255813954, + "aime/avg_subtask_length": 143.35142783637767, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0030303030303030303, + "2": 0.11212121212121212, + "3": 0.831060606060606, + "4": 0.0446969696969697, + "5": 0.00909090909090909 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.040697674418604654, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.004373552868536146, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.26935940313866735, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01775147928994083, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7128891175713918, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.02803030303030303, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9719696969696969, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.41413373860182373, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011398176291793313, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.574468085106383, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3792808219178082, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.039383561643835614, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5813356164383562, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.28169014084507044, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.11267605633802817, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6056338028169014, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8640988372093024, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 439.0, + "completions/max_terminated_length": 439.0, + "completions/mean_length": 186.5390625, + "completions/mean_terminated_length": 186.5390625, + "completions/min_length": 84.0, + "completions/min_terminated_length": 84.0, + "epoch": 0.18627450980392157, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9267241379310345, + "global/avg_subtask_length": 127.66635426429241, + "global/chosen_num_routing_steps_ratio": { + "1": 0.04796511627906977, + "2": 0.18488372093023256, + "3": 0.7104651162790697, + "4": 0.04098837209302326, + "5": 0.015697674418604653 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.07327586206896551, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.020202020202020204, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3774862022284703, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18733729043007394, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4351765073414558, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.40436046511627904, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21686046511627907, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.37877906976744186, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3563358778625954, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.14229007633587787, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5013740458015267, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3774156877605154, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18719211822660098, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.43539219401288365, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3128205128205128, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3282051282051282, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.358974358974359, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18518518518518517, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5370370370370371, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2777777777777778, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7583512931034483, + "global/total_cost": 0.0, + "num_tokens": 15695912.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9442567567567568, + "simpleqa_verified/avg_subtask_length": 99.37235271467077, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.14221824686940965, + "2": 0.3962432915921288, + "3": 0.4588550983899821, + "4": 0.0017889087656529517, + "5": 0.0008944543828264759 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.05574324324324324, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.023488640739314594, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.627262225644975, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04351174432036966, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3292260300346554, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9928443649373881, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0008944543828264759, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.006261180679785331, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15849843587069865, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.035453597497393116, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8060479666319083, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7034883720930233, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1511627906976744, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14534883720930233, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7136824324324325, + "simpleqa_verified/total_cost": 0.0, + "step": 57, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8697916666666666, + "swe_bench/avg_subtask_length": 131.67778134017314, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.001996007984031936, + "2": 0.04491017964071856, + "3": 0.8323353293413174, + "4": 0.07984031936127745, + "5": 0.04091816367265469 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.13020833333333334, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.037191407502404614, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30426418723949983, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5184353959602437, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1773004168002565, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2435129740518962, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7435129740518962, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012974051896207584, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.47, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.417, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.113, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.19895287958115182, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.387434554973822, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41361256544502617, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3140495867768595, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4628099173553719, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2231404958677686, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.17073170731707318, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7073170731707317, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.12195121951219512, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6779513888888888, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3367625791579485, + "epoch": 0.1895424836601307, + "frac_reward_zero_std": 1.0, + "grad_norm": 0.0, + "learning_rate": 8.389541930037516e-07, + "loss": 0.0, + "reward": 1.25, + "reward_std": 0.0, + "rewards/MultiTurnTaskReward/mean": 1.25, + "rewards/MultiTurnTaskReward/std": 0.434714138507843, + "sampling/importance_sampling_ratio/max": 1.9423831701278687, + "sampling/importance_sampling_ratio/mean": 1.0005823373794556, + "sampling/importance_sampling_ratio/min": 0.33724385499954224, + "sampling/sampling_logp_difference/max": 1.086949110031128, + "sampling/sampling_logp_difference/mean": 0.01909565180540085, + "step": 58 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9602272727272727, + "aime/avg_subtask_length": 143.94724943481538, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0029585798816568047, + "2": 0.11094674556213018, + "3": 0.8335798816568047, + "4": 0.04363905325443787, + "5": 0.008875739644970414 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03977272727272727, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.004270283848279327, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.270032655111781, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.017583521728208994, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.71238382316001, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.027366863905325445, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9726331360946746, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4169139465875371, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01112759643916914, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5719584569732937, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3789649415692821, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03923205342237062, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5818030050083473, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.28169014084507044, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.11267605633802817, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6056338028169014, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8671875, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 238.0, + "completions/max_terminated_length": 238.0, + "completions/mean_length": 147.734375, + "completions/mean_terminated_length": 147.734375, + "completions/min_length": 52.0, + "completions/min_terminated_length": 52.0, + "epoch": 0.1895424836601307, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9279661016949152, + "global/avg_subtask_length": 127.66444944740074, + "global/chosen_num_routing_steps_ratio": { + "1": 0.047659817351598174, + "2": 0.18692922374429224, + "3": 0.7097602739726028, + "4": 0.04023972602739726, + "5": 0.015410958904109588 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.07203389830508475, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.019852640196479737, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3791444944740074, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18430208759721653, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4365534179287761, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4061073059360731, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21289954337899544, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3809931506849315, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.35720707222055736, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.13994605933473178, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5028468684447108, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.37994034302759133, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18456375838926176, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4354958985831469, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3128205128205128, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3282051282051282, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.358974358974359, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18518518518518517, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5370370370370371, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2777777777777778, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7621822033898306, + "global/total_cost": 0.0, + "num_tokens": 15888710.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9457236842105263, + "simpleqa_verified/avg_subtask_length": 98.72005988023952, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.14, + "2": 0.4, + "3": 0.4573913043478261, + "4": 0.0017391304347826088, + "5": 0.0008695652173913044 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.054276315789473686, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.02282934131736527, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6291167664670658, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04266467065868264, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3282185628742515, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9930434782608696, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0008695652173913044, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.00608695652173913, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.16177957532861476, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03538928210313448, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8028311425682507, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7088846880907372, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.14744801512287334, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14366729678638943, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7203947368421053, + "simpleqa_verified/total_cost": 0.0, + "step": 58, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8697916666666666, + "swe_bench/avg_subtask_length": 131.67778134017314, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.001996007984031936, + "2": 0.04491017964071856, + "3": 0.8323353293413174, + "4": 0.07984031936127745, + "5": 0.04091816367265469 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.13020833333333334, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.037191407502404614, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30426418723949983, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5184353959602437, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1773004168002565, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2435129740518962, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7435129740518962, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012974051896207584, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.47, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.417, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.113, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.19895287958115182, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.387434554973822, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41361256544502617, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3140495867768595, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4628099173553719, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2231404958677686, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.17073170731707318, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7073170731707317, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.12195121951219512, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6779513888888888, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.30571188777685165, + "epoch": 0.19281045751633988, + "frac_reward_zero_std": 0.5, + "grad_norm": 0.9605126976966858, + "learning_rate": 8.329576125058405e-07, + "loss": 0.0028, + "reward": 1.2265625, + "reward_std": 0.10567784309387207, + "rewards/MultiTurnTaskReward/mean": 1.2265625, + "rewards/MultiTurnTaskReward/std": 0.447474867105484, + "sampling/importance_sampling_ratio/max": 1.8718515634536743, + "sampling/importance_sampling_ratio/mean": 0.9998469352722168, + "sampling/importance_sampling_ratio/min": 0.4848617911338806, + "sampling/sampling_logp_difference/max": 0.7238914370536804, + "sampling/sampling_logp_difference/mean": 0.016357673332095146, + "step": 59 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9611111111111111, + "aime/avg_subtask_length": 143.97621383030898, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.002890173410404624, + "2": 0.10838150289017341, + "3": 0.8367052023121387, + "4": 0.04335260115606936, + "5": 0.008670520231213872 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03888888888888889, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.004168710152035311, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.26998528690534573, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01765571358509073, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7123589995095635, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.026734104046242775, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9732658959537572, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.41304347826086957, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.011594202898550725, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5753623188405798, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.383739837398374, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03902439024390244, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5772357723577236, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2777777777777778, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.1111111111111111, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6111111111111112, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8701388888888889, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 359.0, + "completions/max_terminated_length": 359.0, + "completions/mean_length": 192.6015625, + "completions/mean_terminated_length": 192.6015625, + "completions/min_length": 79.0, + "completions/min_terminated_length": 79.0, + "epoch": 0.19281045751633988, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9291666666666667, + "global/avg_subtask_length": 128.0647813878861, + "global/chosen_num_routing_steps_ratio": { + "1": 0.04680493273542601, + "2": 0.1835762331838565, + "3": 0.7130044843049327, + "4": 0.04119955156950673, + "5": 0.015414798206278026 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.07083333333333333, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.01945447252306458, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.37735659847573205, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1865222623345367, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4361211391897312, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.40106502242152464, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2158071748878924, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.38312780269058294, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.35724786827403704, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.14231108497500736, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5004410467509556, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3798252002913328, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.185360524399126, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.43481427530954114, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31683168316831684, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3316831683168317, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.35148514851485146, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5454545454545454, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2727272727272727, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7598958333333333, + "global/total_cost": 0.0, + "num_tokens": 16135315.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9457236842105263, + "simpleqa_verified/avg_subtask_length": 98.72005988023952, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.14, + "2": 0.4, + "3": 0.4573913043478261, + "4": 0.0017391304347826088, + "5": 0.0008695652173913044 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.054276315789473686, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.02282934131736527, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6291167664670658, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04266467065868264, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3282185628742515, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9930434782608696, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0008695652173913044, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.00608695652173913, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.16177957532861476, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03538928210313448, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8028311425682507, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7088846880907372, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.14744801512287334, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14366729678638943, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7203947368421053, + "simpleqa_verified/total_cost": 0.0, + "step": 59, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8733108108108109, + "swe_bench/avg_subtask_length": 132.26163873370578, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0019342359767891683, + "2": 0.043520309477756286, + "3": 0.8317214700193424, + "4": 0.08220502901353965, + "5": 0.04061895551257253 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.1266891891891892, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.03600248292985723, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30446927374301674, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5195530726256983, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17597765363128492, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2437137330754352, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7437137330754352, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012572533849129593, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4699612403100775, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.41957364341085274, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.11046511627906977, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.19858156028368795, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3880445795339412, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4133738601823708, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3228346456692913, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4645669291338583, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2125984251968504, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11904761904761904, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6663851351351351, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.3262897692620754, + "epoch": 0.19607843137254902, + "frac_reward_zero_std": 0.5, + "grad_norm": 1.0161094665527344, + "learning_rate": 8.268737196446263e-07, + "loss": -0.0065, + "reward": 1.203125, + "reward_std": 0.23598936200141907, + "rewards/MultiTurnTaskReward/mean": 1.203125, + "rewards/MultiTurnTaskReward/std": 0.45867621898651123, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.000377893447876, + "sampling/importance_sampling_ratio/min": 0.15809059143066406, + "sampling/sampling_logp_difference/max": 1.8445870876312256, + "sampling/sampling_logp_difference/mean": 0.017682891339063644, + "step": 60 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9619565217391305, + "aime/avg_subtask_length": 144.50587389115321, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.002824858757062147, + "2": 0.10805084745762712, + "3": 0.838276836158192, + "4": 0.0423728813559322, + "5": 0.00847457627118644 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03804347826086957, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.004075761208343323, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2709182450251738, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.017262047470630544, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7118197075041957, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.026129943502824857, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9738700564971752, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4128895184135977, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0113314447592068, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5757790368271954, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3876092136616362, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03812549642573471, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.574265289912629, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2777777777777778, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.1111111111111111, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6111111111111112, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8729619565217391, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 419.0, + "completions/max_terminated_length": 419.0, + "completions/mean_length": 190.65625, + "completions/mean_terminated_length": 190.65625, + "completions/min_length": 104.0, + "completions/min_terminated_length": 104.0, + "epoch": 0.19607843137254902, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.930327868852459, + "global/avg_subtask_length": 128.47022930813895, + "global/chosen_num_routing_steps_ratio": { + "1": 0.04598017621145375, + "2": 0.18116740088105726, + "3": 0.7172356828193832, + "4": 0.04047356828193833, + "5": 0.0151431718061674 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.06967213114754098, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.01909260899517764, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3765377423481941, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1875799626021061, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.43588229504969983, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3986784140969163, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21613436123348018, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.38518722466960353, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.35613275613275613, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1471861471861472, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49668109668109667, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3811898824367652, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18311364446027786, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4356964731029569, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31683168316831684, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3316831683168317, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.35148514851485146, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5454545454545454, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2727272727272727, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7635758196721312, + "global/total_cost": 0.0, + "num_tokens": 16436103.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9457236842105263, + "simpleqa_verified/avg_subtask_length": 98.72005988023952, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.14, + "2": 0.4, + "3": 0.4573913043478261, + "4": 0.0017391304347826088, + "5": 0.0008695652173913044 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.054276315789473686, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.02282934131736527, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6291167664670658, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04266467065868264, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3282185628742515, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9930434782608696, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0008695652173913044, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.00608695652173913, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.16177957532861476, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03538928210313448, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8028311425682507, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7088846880907372, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.14744801512287334, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14366729678638943, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7203947368421053, + "simpleqa_verified/total_cost": 0.0, + "step": 60, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8766447368421053, + "swe_bench/avg_subtask_length": 132.27004219409284, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.001876172607879925, + "2": 0.04221388367729831, + "3": 0.8367729831144465, + "4": 0.07973733583489681, + "5": 0.039399624765478425 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.12335526315789473, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.03496081977094635, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3059071729957806, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5183845690174804, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.175708257986739, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2523452157598499, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7354596622889306, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012195121951219513, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.46146616541353386, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.43139097744360905, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.10714285714285714, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.20314033366045142, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.380765456329735, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41609421000981356, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3228346456692913, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4645669291338583, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2125984251968504, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11904761904761904, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6743421052631579, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.37395663745701313, + "epoch": 0.19934640522875818, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.4990101158618927, + "learning_rate": 8.207041098155699e-07, + "loss": -0.0188, + "reward": 1.2421875, + "reward_std": 0.04419417306780815, + "rewards/MultiTurnTaskReward/mean": 1.2421875, + "rewards/MultiTurnTaskReward/std": 0.43914902210235596, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0006353855133057, + "sampling/importance_sampling_ratio/min": 0.06994962692260742, + "sampling/sampling_logp_difference/max": 2.659979820251465, + "sampling/sampling_logp_difference/mean": 0.020083710551261902, + "step": 61 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9627659574468085, + "aime/avg_subtask_length": 144.7149554617909, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0027624309392265192, + "2": 0.106353591160221, + "3": 0.8411602209944752, + "4": 0.04143646408839779, + "5": 0.008287292817679558 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03723404255319149, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0039849976558837315, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2712142522269105, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01711204875761838, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7116736990154712, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.025552486187845305, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9744475138121547, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4120498614958449, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0110803324099723, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5768698060941828, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.38992248062015505, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03798449612403101, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5720930232558139, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2777777777777778, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.1111111111111111, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6111111111111112, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8756648936170213, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 313.0, + "completions/max_terminated_length": 313.0, + "completions/mean_length": 183.53125, + "completions/mean_terminated_length": 183.53125, + "completions/min_length": 91.0, + "completions/min_terminated_length": 91.0, + "epoch": 0.19934640522875818, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9314516129032258, + "global/avg_subtask_length": 128.7986674391657, + "global/chosen_num_routing_steps_ratio": { + "1": 0.04518398268398268, + "2": 0.1783008658008658, + "3": 0.7213203463203464, + "4": 0.03977272727272727, + "5": 0.015422077922077922 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.06854838709677419, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.01873310158362302, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.37514484356894556, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18935882580146776, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4354963306295867, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.3963744588744589, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21645021645021645, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3871753246753247, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3547747237177671, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1518843865117597, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49334088977047325, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3808362369337979, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18327526132404182, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.43588850174216026, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3137254901960784, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3333333333333333, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.35294117647058826, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.19298245614035087, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.543859649122807, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2631578947368421, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7673891129032258, + "global/total_cost": 0.0, + "num_tokens": 16687915.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9457236842105263, + "simpleqa_verified/avg_subtask_length": 98.72005988023952, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.14, + "2": 0.4, + "3": 0.4573913043478261, + "4": 0.0017391304347826088, + "5": 0.0008695652173913044 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.054276315789473686, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.02282934131736527, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6291167664670658, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04266467065868264, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3282185628742515, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9930434782608696, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0008695652173913044, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.00608695652173913, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.16177957532861476, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03538928210313448, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8028311425682507, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7088846880907372, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.14744801512287334, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14366729678638943, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7203947368421053, + "simpleqa_verified/total_cost": 0.0, + "step": 61, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8798076923076923, + "swe_bench/avg_subtask_length": 132.4473376243417, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0018214936247723133, + "2": 0.040983606557377046, + "3": 0.8397085610200364, + "4": 0.07741347905282331, + "5": 0.04007285974499089 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.1201923076923077, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.03393797542422469, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30631948507899354, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5190169689877121, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17466354593329433, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2604735883424408, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7276867030965392, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011839708561020037, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4534671532846715, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4425182481751825, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.10401459854014598, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.20456707897240722, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3796384395813511, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41579448144624165, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3178294573643411, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46511627906976744, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.21705426356589147, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7045454545454546, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11363636363636363, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6826923076923077, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.29922492429614067, + "epoch": 0.20261437908496732, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.5768920183181763, + "learning_rate": 8.144504008919222e-07, + "loss": 0.0109, + "reward": 1.33203125, + "reward_std": 0.12544025480747223, + "rewards/MultiTurnTaskReward/mean": 1.33203125, + "rewards/MultiTurnTaskReward/std": 0.38309425115585327, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0001256465911865, + "sampling/importance_sampling_ratio/min": 0.35887610912323, + "sampling/sampling_logp_difference/max": 1.024778127670288, + "sampling/sampling_logp_difference/mean": 0.016340889036655426, + "step": 62 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9627659574468085, + "aime/avg_subtask_length": 144.7149554617909, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0027624309392265192, + "2": 0.106353591160221, + "3": 0.8411602209944752, + "4": 0.04143646408839779, + "5": 0.008287292817679558 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03723404255319149, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0039849976558837315, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2712142522269105, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01711204875761838, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7116736990154712, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.025552486187845305, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9744475138121547, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4120498614958449, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0110803324099723, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5768698060941828, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.38992248062015505, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03798449612403101, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5720930232558139, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2777777777777778, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.1111111111111111, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6111111111111112, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8756648936170213, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 366.0, + "completions/max_terminated_length": 366.0, + "completions/mean_length": 199.1328125, + "completions/mean_terminated_length": 199.1328125, + "completions/min_length": 73.0, + "completions/min_terminated_length": 73.0, + "epoch": 0.20261437908496732, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9325396825396826, + "global/avg_subtask_length": 128.6279202279202, + "global/chosen_num_routing_steps_ratio": { + "1": 0.0449468085106383, + "2": 0.17952127659574468, + "3": 0.7207446808510638, + "4": 0.03962765957446809, + "5": 0.015159574468085106 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.06746031746031746, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.01842355175688509, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3762583095916429, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19040835707502374, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.43333333333333335, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4018617021276596, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2170212765957447, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.38111702127659575, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3517126148705096, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1542745753272069, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49401280980228346, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3813443072702332, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1834705075445816, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4351851851851852, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3155339805825243, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.33495145631067963, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.34951456310679613, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.19298245614035087, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.543859649122807, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2631578947368421, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7710813492063492, + "global/total_cost": 0.0, + "num_tokens": 16970684.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9471153846153846, + "simpleqa_verified/avg_subtask_length": 98.14592430858806, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.13790186125211507, + "2": 0.4027072758037225, + "3": 0.45685279187817257, + "4": 0.001692047377326565, + "5": 0.0008460236886632825 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.052884615384615384, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.02219796215429403, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6299126637554585, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.041848617176128096, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3282387190684134, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9932318104906938, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0008460236886632825, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.005922165820642978, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.1619234543670265, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03434739941118744, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.803729146221786, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.714548802946593, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.14548802946593, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.13996316758747698, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7275641025641025, + "simpleqa_verified/total_cost": 0.0, + "step": 62, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8828125, + "swe_bench/avg_subtask_length": 132.93316268486916, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0017699115044247787, + "2": 0.03982300884955752, + "3": 0.8424778761061947, + "4": 0.07699115044247788, + "5": 0.03893805309734513 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.1171875, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.03299203640500569, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3054607508532423, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5167804323094426, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17775881683731512, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.26548672566371684, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7212389380530974, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.01327433628318584, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.44592198581560283, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.44592198581560283, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.10815602836879433, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.20406278855032317, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.37580794090489383, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.420129270544783, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.32061068702290074, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46564885496183206, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.21374045801526717, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7045454545454546, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11363636363636363, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.690625, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.278933634981513, + "epoch": 0.20588235294117646, + "frac_reward_zero_std": 1.0, + "grad_norm": 0.0, + "learning_rate": 8.081142328004636e-07, + "loss": 0.0, + "reward": 1.25, + "reward_std": 0.0, + "rewards/MultiTurnTaskReward/mean": 1.25, + "rewards/MultiTurnTaskReward/std": 0.434714138507843, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9997833967208862, + "sampling/importance_sampling_ratio/min": 0.17946745455265045, + "sampling/sampling_logp_difference/max": 1.7177613973617554, + "sampling/sampling_logp_difference/mean": 0.016966458410024643, + "step": 63 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9635416666666666, + "aime/avg_subtask_length": 145.17416245984396, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.002702702702702703, + "2": 0.10675675675675676, + "3": 0.8418918918918918, + "4": 0.04054054054054054, + "5": 0.008108108108108109 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.036458333333333336, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.003900871959614502, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2721431849472235, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.016750803120697568, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7111060119320789, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.025, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.975, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.41192411924119243, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01084010840108401, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5772357723577236, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3937784522003035, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03717754172989378, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5690440060698028, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2777777777777778, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.1111111111111111, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6111111111111112, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8782552083333334, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 497.0, + "completions/max_terminated_length": 497.0, + "completions/mean_length": 187.28125, + "completions/mean_terminated_length": 187.28125, + "completions/min_length": 107.0, + "completions/min_terminated_length": 107.0, + "epoch": 0.20588235294117646, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.93359375, + "global/avg_subtask_length": 128.75490013067014, + "global/chosen_num_routing_steps_ratio": { + "1": 0.04419456066945607, + "2": 0.17860878661087867, + "3": 0.7233263598326359, + "4": 0.038964435146443516, + "5": 0.014905857740585775 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.06640625, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.018107149523987308, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.37791674444651857, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1872316595109203, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.43485159604256113, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4035041841004184, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21338912133891214, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3831066945606695, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3507523939808482, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.15157318741450068, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49767441860465117, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.38627187079407804, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.180349932705249, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.43337819650067294, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3155339805825243, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.33495145631067963, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.34951456310679613, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.19298245614035087, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.543859649122807, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2631578947368421, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.774658203125, + "global/total_cost": 0.0, + "num_tokens": 17215328.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9484375, + "simpleqa_verified/avg_subtask_length": 98.38661971830986, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.1342668863261944, + "2": 0.3953871499176277, + "3": 0.46787479406919275, + "4": 0.0016474464579901153, + "5": 0.0008237232289950577 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.0515625, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.02147887323943662, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6299295774647887, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04084507042253521, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.32922535211267606, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9934102141680395, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0008237232289950577, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.005766062602965404, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.16270218839200762, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03330161750713606, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8039961941008563, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7145359019264448, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.14010507880910683, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14535901926444833, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.734375, + "simpleqa_verified/total_cost": 0.0, + "step": 63, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8828125, + "swe_bench/avg_subtask_length": 132.93316268486916, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0017699115044247787, + "2": 0.03982300884955752, + "3": 0.8424778761061947, + "4": 0.07699115044247788, + "5": 0.03893805309734513 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.1171875, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.03299203640500569, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3054607508532423, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5167804323094426, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17775881683731512, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.26548672566371684, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7212389380530974, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.01327433628318584, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.44592198581560283, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.44592198581560283, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.10815602836879433, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.20406278855032317, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.37580794090489383, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.420129270544783, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.32061068702290074, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46564885496183206, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.21374045801526717, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7045454545454546, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11363636363636363, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.690625, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.2610591808333993, + "epoch": 0.20915032679738563, + "frac_reward_zero_std": 0.5, + "grad_norm": 0.9685524106025696, + "learning_rate": 8.016972670914623e-07, + "loss": -0.0075, + "reward": 1.25, + "reward_std": 0.1480722874403, + "rewards/MultiTurnTaskReward/mean": 1.25, + "rewards/MultiTurnTaskReward/std": 0.434714138507843, + "sampling/importance_sampling_ratio/max": 1.773108720779419, + "sampling/importance_sampling_ratio/mean": 1.0000760555267334, + "sampling/importance_sampling_ratio/min": 0.41616448760032654, + "sampling/sampling_logp_difference/max": 0.8766746520996094, + "sampling/sampling_logp_difference/mean": 0.015295442193746567, + "step": 64 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9635416666666666, + "aime/avg_subtask_length": 145.17416245984396, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.002702702702702703, + "2": 0.10675675675675676, + "3": 0.8418918918918918, + "4": 0.04054054054054054, + "5": 0.008108108108108109 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.036458333333333336, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.003900871959614502, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2721431849472235, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.016750803120697568, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7111060119320789, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.025, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.975, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.41192411924119243, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01084010840108401, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5772357723577236, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3937784522003035, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03717754172989378, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5690440060698028, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2777777777777778, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.1111111111111111, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6111111111111112, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8782552083333334, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 459.0, + "completions/max_terminated_length": 459.0, + "completions/mean_length": 223.6640625, + "completions/mean_terminated_length": 223.6640625, + "completions/min_length": 96.0, + "completions/min_terminated_length": 96.0, + "epoch": 0.20915032679738563, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9346153846153846, + "global/avg_subtask_length": 129.0582141547488, + "global/chosen_num_routing_steps_ratio": { + "1": 0.043467078189300415, + "2": 0.17618312757201646, + "3": 0.7263374485596708, + "4": 0.03935185185185185, + "5": 0.014660493827160493 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.06538461538461539, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.017785111844517785, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.37642097543087644, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19416941694169418, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4294096076274294, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4017489711934156, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22145061728395063, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.37680041152263377, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3492874428609841, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.15998924442054316, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4907233127184727, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3849703361898484, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1845748187211602, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.43045484508899146, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3142857142857143, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3380952380952381, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3476190476190476, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.19298245614035087, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.543859649122807, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2631578947368421, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7776442307692307, + "global/total_cost": 0.0, + "num_tokens": 17538709.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9484375, + "simpleqa_verified/avg_subtask_length": 98.38661971830986, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.1342668863261944, + "2": 0.3953871499176277, + "3": 0.46787479406919275, + "4": 0.0016474464579901153, + "5": 0.0008237232289950577 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.0515625, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.02147887323943662, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6299295774647887, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.04084507042253521, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.32922535211267606, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9934102141680395, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0008237232289950577, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.005766062602965404, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.16270218839200762, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03330161750713606, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8039961941008563, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7145359019264448, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.14010507880910683, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14535901926444833, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.734375, + "simpleqa_verified/total_cost": 0.0, + "step": 64, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8883928571428571, + "swe_bench/avg_subtask_length": 133.6064690026954, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0016750418760469012, + "2": 0.03936348408710218, + "3": 0.8458961474036851, + "4": 0.076214405360134, + "5": 0.03685092127303183 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.11160714285714286, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.031266846361185985, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30485175202156334, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5199460916442048, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1752021563342318, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.26716917922948075, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7202680067001676, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.01256281407035176, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.436241610738255, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4563758389261745, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.10738255033557047, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.21048034934497817, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3764192139737991, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4131004366812227, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31851851851851853, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4666666666666667, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.21481481481481482, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7045454545454546, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11363636363636363, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.7038690476190477, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.24030884820967913, + "epoch": 0.21241830065359477, + "frac_reward_zero_std": 0.5, + "grad_norm": 0.849739134311676, + "learning_rate": 7.952011865029613e-07, + "loss": -0.0044, + "reward": 1.4765625, + "reward_std": 0.10567784309387207, + "rewards/MultiTurnTaskReward/mean": 1.4765625, + "rewards/MultiTurnTaskReward/std": 0.15188287198543549, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9996759295463562, + "sampling/importance_sampling_ratio/min": 0.09877597540616989, + "sampling/sampling_logp_difference/max": 2.3149008750915527, + "sampling/sampling_logp_difference/mean": 0.017416000366210938, + "step": 65 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9635416666666666, + "aime/avg_subtask_length": 145.17416245984396, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.002702702702702703, + "2": 0.10675675675675676, + "3": 0.8418918918918918, + "4": 0.04054054054054054, + "5": 0.008108108108108109 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.036458333333333336, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.003900871959614502, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 15, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2721431849472235, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.016750803120697568, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7111060119320789, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.025, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.975, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.41192411924119243, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01084010840108401, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5772357723577236, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3937784522003035, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03717754172989378, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5690440060698028, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.2777777777777778, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.1111111111111111, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.6111111111111112, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8782552083333334, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 514.0, + "completions/max_terminated_length": 514.0, + "completions/mean_length": 227.0078125, + "completions/mean_terminated_length": 227.0078125, + "completions/min_length": 113.0, + "completions/min_terminated_length": 113.0, + "epoch": 0.21241830065359477, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9356060606060606, + "global/avg_subtask_length": 129.04426613775695, + "global/chosen_num_routing_steps_ratio": { + "1": 0.04276315789473684, + "2": 0.17586032388663966, + "3": 0.7277327935222672, + "4": 0.039220647773279355, + "5": 0.014423076923076924 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.06439393939393939, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.017490082942661377, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 15, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3780201947349441, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19527587450414713, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4267039307609088, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.40789473684210525, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22140688259109312, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3706983805668016, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3465503568596352, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1633624107850912, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49008723235527357, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3856865284974093, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18490932642487046, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.42940414507772023, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.32075471698113206, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.33490566037735847, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3443396226415094, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.19298245614035087, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.543859649122807, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2631578947368421, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7677556818181818, + "global/total_cost": 0.0, + "num_tokens": 17977302.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9496951219512195, + "simpleqa_verified/avg_subtask_length": 98.62487188247353, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.13081861958266452, + "2": 0.3924558587479936, + "3": 0.47431781701444625, + "4": 0.0016051364365971107, + "5": 0.0008025682182985554 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.05030487804878049, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.020840450973693202, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6299965835326272, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.040655961735565425, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3293474547318073, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9935794542536116, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0008025682182985554, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.0056179775280898875, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.1634349030470914, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0332409972299169, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8033240997229917, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7154882154882155, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.13804713804713806, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14646464646464646, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7164634146341463, + "simpleqa_verified/total_cost": 0.0, + "step": 65, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8909883720930233, + "swe_bench/avg_subtask_length": 133.96769109535066, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0016313213703099511, + "2": 0.03915171288743882, + "3": 0.8474714518760196, + "4": 0.07585644371941272, + "5": 0.03588907014681892 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.10901162790697674, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.030470186498555294, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3054898870501707, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5185185185185185, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17599159443131074, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.27487765089722677, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7128874388254486, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012234910277324634, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4297385620915033, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4624183006535948, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.10784313725490197, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2100340136054422, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3741496598639456, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41581632653061223, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3284671532846715, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.45985401459854014, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2116788321167883, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7045454545454546, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11363636363636363, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6933139534883721, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.2765926318243146, + "epoch": 0.21568627450980393, + "frac_reward_zero_std": 0.5, + "grad_norm": 1.002493977546692, + "learning_rate": 7.886276945195097e-07, + "loss": 0.0114, + "reward": 0.921875, + "reward_std": 0.23598936200141907, + "rewards/MultiTurnTaskReward/mean": 0.921875, + "rewards/MultiTurnTaskReward/std": 0.4957992732524872, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9997760057449341, + "sampling/importance_sampling_ratio/min": 0.029154159128665924, + "sampling/sampling_logp_difference/max": 3.5351576805114746, + "sampling/sampling_logp_difference/mean": 0.017523441463708878, + "step": 66 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9642857142857143, + "aime/avg_subtask_length": 145.6690599057662, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0026455026455026454, + "2": 0.10515873015873016, + "3": 0.841931216931217, + "4": 0.042328042328042326, + "5": 0.007936507936507936 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03571428571428571, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.003814224814897913, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 16, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2728292573479919, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.016603096253085033, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.710567646398923, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.02447089947089947, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9755291005291006, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.40981432360742703, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.010610079575596816, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5795755968169761, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3973313565604151, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.037064492216456635, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5656041512231282, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3026315789473684, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.10526315789473684, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5921052631578947, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8807397959183674, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 389.0, + "completions/max_terminated_length": 389.0, + "completions/mean_length": 207.140625, + "completions/mean_terminated_length": 207.140625, + "completions/min_length": 70.0, + "completions/min_terminated_length": 70.0, + "epoch": 0.21568627450980393, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9365671641791045, + "global/avg_subtask_length": 129.1885035039475, + "global/chosen_num_routing_steps_ratio": { + "1": 0.042081673306772906, + "2": 0.17679282868525897, + "3": 0.7273406374501992, + "4": 0.03959163346613546, + "5": 0.014193227091633466 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.06343283582089553, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.017209261066264525, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 16, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.379135988645436, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19222922026080014, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.42863479109376385, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.40936254980079684, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2178784860557769, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3727589641434263, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.34364439823238885, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.16064465817520146, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4957109435924097, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3908192540643927, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18233981510997768, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.42684093082562957, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3287037037037037, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3287037037037037, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3425925925925926, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.19298245614035087, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.543859649122807, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2631578947368421, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7709888059701493, + "global/total_cost": 0.0, + "num_tokens": 18223592.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9508928571428571, + "simpleqa_verified/avg_subtask_length": 98.73047524094383, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.12754303599374023, + "2": 0.3935837245696401, + "3": 0.4765258215962441, + "4": 0.001564945226917058, + "5": 0.000782472613458529 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.049107142857142856, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.020272515785975408, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6297773346626786, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.03954802259887006, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3306746427384513, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9937402190923318, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.000782472613458529, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.005477308294209703, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15964125560538117, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03228699551569507, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8080717488789237, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7238562091503268, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.13398692810457516, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14215686274509803, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7224702380952381, + "simpleqa_verified/total_cost": 0.0, + "step": 66, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8909883720930233, + "swe_bench/avg_subtask_length": 133.96769109535066, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0016313213703099511, + "2": 0.03915171288743882, + "3": 0.8474714518760196, + "4": 0.07585644371941272, + "5": 0.03588907014681892 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.10901162790697674, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.030470186498555294, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3054898870501707, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5185185185185185, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17599159443131074, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.27487765089722677, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7128874388254486, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012234910277324634, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4297385620915033, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4624183006535948, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.10784313725490197, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2100340136054422, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3741496598639456, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41581632653061223, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3284671532846715, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.45985401459854014, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2116788321167883, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7045454545454546, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11363636363636363, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6933139534883721, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.2697121184319258, + "epoch": 0.21895424836601307, + "frac_reward_zero_std": 0.25, + "grad_norm": 1.108373761177063, + "learning_rate": 7.819785149254532e-07, + "loss": 0.0116, + "reward": 1.328125, + "reward_std": 0.23042556643486023, + "rewards/MultiTurnTaskReward/mean": 1.328125, + "rewards/MultiTurnTaskReward/std": 0.3787541687488556, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9996431469917297, + "sampling/importance_sampling_ratio/min": 0.12188265472650528, + "sampling/sampling_logp_difference/max": 2.10469651222229, + "sampling/sampling_logp_difference/mean": 0.016501259058713913, + "step": 67 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.965, + "aime/avg_subtask_length": 145.84713375796179, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0025906735751295338, + "2": 0.10492227979274611, + "3": 0.841321243523316, + "4": 0.04339378238341969, + "5": 0.007772020725388601 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.035, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0037338018888644848, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 16, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2725675378871074, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.016253019986821875, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7111794421260708, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.023963730569948185, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9760362694300518, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.40844155844155844, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01038961038961039, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5811688311688312, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3969521044992743, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.036284470246734396, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5667634252539913, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3291139240506329, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.10126582278481013, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.569620253164557, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.883125, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 359.0, + "completions/max_terminated_length": 359.0, + "completions/mean_length": 161.1328125, + "completions/mean_terminated_length": 161.1328125, + "completions/min_length": 82.0, + "completions/min_terminated_length": 82.0, + "epoch": 0.21895424836601307, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9375, + "global/avg_subtask_length": 129.24978165938865, + "global/chosen_num_routing_steps_ratio": { + "1": 0.04142156862745098, + "2": 0.1784313725490196, + "3": 0.7264705882352941, + "4": 0.039705882352941174, + "5": 0.013970588235294118 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.0625, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.016943231441048034, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 16, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3799126637554585, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.18925764192139738, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4308296943231441, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4107843137254902, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.21446078431372548, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3747549019607843, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.34160061365379696, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1580158527230887, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5003835336231143, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3936537857367264, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.17970468111844173, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.42664153314483194, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3378995433789954, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3242009132420091, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3378995433789954, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.19298245614035087, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.543859649122807, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2631578947368421, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7670036764705882, + "global/total_cost": 0.0, + "num_tokens": 18420505.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9520348837209303, + "simpleqa_verified/avg_subtask_length": 98.98155339805825, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.12442748091603054, + "2": 0.39541984732824426, + "3": 0.4778625954198473, + "4": 0.0015267175572519084, + "5": 0.0007633587786259542 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.04796511627906977, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.019741100323624597, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6297734627831715, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.038511326860841424, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.33171521035598706, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9938931297709923, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0007633587786259542, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.0053435114503816794, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15780296425457715, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03138622493461203, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8108108108108109, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7297297297297297, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.13036565977742448, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.13990461049284578, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7056686046511628, + "simpleqa_verified/total_cost": 0.0, + "step": 67, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8909883720930233, + "swe_bench/avg_subtask_length": 133.96769109535066, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0016313213703099511, + "2": 0.03915171288743882, + "3": 0.8474714518760196, + "4": 0.07585644371941272, + "5": 0.03588907014681892 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.10901162790697674, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.030470186498555294, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3054898870501707, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5185185185185185, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17599159443131074, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.27487765089722677, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7128874388254486, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.012234910277324634, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4297385620915033, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4624183006535948, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.10784313725490197, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2100340136054422, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3741496598639456, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41581632653061223, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3284671532846715, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.45985401459854014, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2116788321167883, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18181818181818182, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7045454545454546, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.11363636363636363, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6933139534883721, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.27180955931544304, + "epoch": 0.2222222222222222, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.6862062811851501, + "learning_rate": 7.752553913529018e-07, + "loss": -0.009, + "reward": 1.2265625, + "reward_std": 0.07403614372015, + "rewards/MultiTurnTaskReward/mean": 1.2265625, + "rewards/MultiTurnTaskReward/std": 0.447474867105484, + "sampling/importance_sampling_ratio/max": 1.5426779985427856, + "sampling/importance_sampling_ratio/mean": 0.9994427561759949, + "sampling/importance_sampling_ratio/min": 0.49984222650527954, + "sampling/sampling_logp_difference/max": 0.6934628486633301, + "sampling/sampling_logp_difference/mean": 0.015356418676674366, + "step": 68 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.965, + "aime/avg_subtask_length": 145.84713375796179, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0025906735751295338, + "2": 0.10492227979274611, + "3": 0.841321243523316, + "4": 0.04339378238341969, + "5": 0.007772020725388601 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.035, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0037338018888644848, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 16, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2725675378871074, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.016253019986821875, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7111794421260708, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.023963730569948185, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9760362694300518, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.40844155844155844, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.01038961038961039, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5811688311688312, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3969521044992743, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.036284470246734396, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5667634252539913, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3291139240506329, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.10126582278481013, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.569620253164557, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.883125, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 443.0, + "completions/max_terminated_length": 443.0, + "completions/mean_length": 199.390625, + "completions/mean_terminated_length": 199.390625, + "completions/min_length": 92.0, + "completions/min_terminated_length": 92.0, + "epoch": 0.2222222222222222, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9384057971014492, + "global/avg_subtask_length": 129.71207828998197, + "global/chosen_num_routing_steps_ratio": { + "1": 0.040781853281853284, + "2": 0.17615830115830117, + "3": 0.7282818532818532, + "4": 0.040781853281853284, + "5": 0.013996138996138996 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.06159420289855073, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.01665379002489484, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 16, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3779723581423298, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19709846338741524, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.42492917847025496, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4075772200772201, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22345559845559845, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3689671814671815, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.33962264150943394, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.16754716981132076, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4928301886792453, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.39322033898305087, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1839753466872111, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4228043143297381, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3392070484581498, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3303964757709251, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3303964757709251, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1896551724137931, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5517241379310345, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.25862068965517243, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7683423913043478, + "global/total_cost": 0.0, + "num_tokens": 18697483.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9520348837209303, + "simpleqa_verified/avg_subtask_length": 98.98155339805825, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.12442748091603054, + "2": 0.39541984732824426, + "3": 0.4778625954198473, + "4": 0.0015267175572519084, + "5": 0.0007633587786259542 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.04796511627906977, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.019741100323624597, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6297734627831715, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.038511326860841424, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.33171521035598706, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9938931297709923, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0007633587786259542, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.0053435114503816794, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15780296425457715, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03138622493461203, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8108108108108109, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7297297297297297, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.13036565977742448, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.13990461049284578, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7056686046511628, + "simpleqa_verified/total_cost": 0.0, + "step": 68, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8958333333333334, + "swe_bench/avg_subtask_length": 135.07763354967548, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0015503875968992248, + "2": 0.03875968992248062, + "3": 0.8472868217054264, + "4": 0.07751937984496124, + "5": 0.03488372093023256 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.10416666666666667, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.02895656515227159, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3035446829755367, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5249625561657514, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17149276085871193, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2713178294573643, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7170542635658915, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011627906976744186, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4192546583850932, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4767080745341615, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.10403726708074534, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.21809369951534732, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3756058158319871, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4063004846526656, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3310344827586207, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46206896551724136, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20689655172413793, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.17777777777777778, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7111111111111111, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.1111111111111111, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.7006944444444444, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.22483266610652208, + "epoch": 0.22549019607843138, + "frac_reward_zero_std": 0.5, + "grad_norm": 1.0072704553604126, + "learning_rate": 7.684600868244919e-07, + "loss": 0.018, + "reward": 1.421875, + "reward_std": 0.15839502215385437, + "rewards/MultiTurnTaskReward/mean": 1.421875, + "rewards/MultiTurnTaskReward/std": 0.2694226801395416, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0002343654632568, + "sampling/importance_sampling_ratio/min": 0.06583281606435776, + "sampling/sampling_logp_difference/max": 2.7206368446350098, + "sampling/sampling_logp_difference/mean": 0.015814730897545815, + "step": 69 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9656862745098039, + "aime/avg_subtask_length": 146.19707085935818, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0025380710659898475, + "2": 0.1065989847715736, + "3": 0.8407360406091371, + "4": 0.04251269035532995, + "5": 0.007614213197969543 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03431372549019608, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.003661425802283007, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 16, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2730992892526384, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01593797113934956, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.710962739608012, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.023477157360406092, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9765228426395939, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.41030534351145037, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.010178117048346057, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5795165394402035, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3974358974358974, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03561253561253561, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5669515669515669, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3291139240506329, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.10126582278481013, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.569620253164557, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8854166666666666, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 422.0, + "completions/max_terminated_length": 422.0, + "completions/mean_length": 204.3515625, + "completions/mean_terminated_length": 204.3515625, + "completions/min_length": 78.0, + "completions/min_terminated_length": 78.0, + "epoch": 0.22549019607843138, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9392857142857143, + "global/avg_subtask_length": 130.18702483527622, + "global/chosen_num_routing_steps_ratio": { + "1": 0.04016159695817491, + "2": 0.17490494296577946, + "3": 0.7302756653992395, + "4": 0.0408745247148289, + "5": 0.013783269961977186 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.060714285714285714, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.016387903362054403, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 16, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.37616151376921775, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19834431491806048, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.42549417131272177, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4056558935361217, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22338403041825095, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.37096007604562736, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.33869769744986383, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.17182470908640754, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.48947759346372866, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3902512867090524, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18346957311534967, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4262791401755979, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3391304347826087, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.33043478260869563, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.33043478260869563, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1896551724137931, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5517241379310345, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.25862068965517243, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7716517857142857, + "global/total_cost": 0.0, + "num_tokens": 18992280.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9520348837209303, + "simpleqa_verified/avg_subtask_length": 98.98155339805825, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.12442748091603054, + "2": 0.39541984732824426, + "3": 0.4778625954198473, + "4": 0.0015267175572519084, + "5": 0.0007633587786259542 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.04796511627906977, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.019741100323624597, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6297734627831715, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.038511326860841424, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.33171521035598706, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9938931297709923, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0007633587786259542, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.0053435114503816794, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15780296425457715, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.03138622493461203, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8108108108108109, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7297297297297297, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.13036565977742448, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.13990461049284578, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7056686046511628, + "simpleqa_verified/total_cost": 0.0, + "step": 69, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8980978260869565, + "swe_bench/avg_subtask_length": 135.56833130328866, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0015128593040847202, + "2": 0.037821482602118005, + "3": 0.848714069591528, + "4": 0.07791225416036308, + "5": 0.0340393343419062 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.10190217391304347, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.02825822168087698, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30182704019488427, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5249695493300852, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17320341047503046, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2783661119515885, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.710287443267776, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.0113464447806354, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4106060606060606, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4863636363636364, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.10303030303030303, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2141732283464567, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3732283464566929, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4125984251968504, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3310810810810811, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4594594594594595, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20945945945945946, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.17777777777777778, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7111111111111111, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.1111111111111111, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.7072010869565217, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.2701421147212386, + "epoch": 0.22875816993464052, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.650816798210144, + "learning_rate": 7.61594383291065e-07, + "loss": -0.01, + "reward": 1.265625, + "reward_std": 0.06148367002606392, + "rewards/MultiTurnTaskReward/mean": 1.265625, + "rewards/MultiTurnTaskReward/std": 0.42527204751968384, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.000063419342041, + "sampling/importance_sampling_ratio/min": 0.37538471817970276, + "sampling/sampling_logp_difference/max": 1.367248296737671, + "sampling/sampling_logp_difference/mean": 0.01613636687397957, + "step": 70 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9663461538461539, + "aime/avg_subtask_length": 146.4731954411144, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0024875621890547263, + "2": 0.10634328358208955, + "3": 0.8407960199004975, + "4": 0.04291044776119403, + "5": 0.007462686567164179 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03365384615384615, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.003588011819333052, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 16, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2728999577880962, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01561840439003799, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7114816378218658, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.023009950248756218, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9769900497512438, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4096009975062344, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.00997506234413965, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.580423940149626, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.39706908583391487, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.034891835310537335, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5680390788555478, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.345679012345679, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.09876543209876543, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5555555555555556, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8876201923076923, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 263.0, + "completions/max_terminated_length": 263.0, + "completions/mean_length": 153.0234375, + "completions/mean_terminated_length": 153.0234375, + "completions/min_length": 85.0, + "completions/min_terminated_length": 85.0, + "epoch": 0.22875816993464052, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9401408450704225, + "global/avg_subtask_length": 130.14395074055582, + "global/chosen_num_routing_steps_ratio": { + "1": 0.03955992509363296, + "2": 0.175561797752809, + "3": 0.7305711610486891, + "4": 0.04073033707865169, + "5": 0.013576779026217229 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.05985915492957746, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.016142452987185887, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 16, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3767681810617407, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19537360625728076, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4278582126809785, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4070692883895131, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2200374531835206, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3728932584269663, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3368267121618328, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.16914452839385816, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.494028759444309, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3924843423799583, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18073367133909932, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.42678198628094244, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3448275862068966, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3275862068965517, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3275862068965517, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1896551724137931, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5517241379310345, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.25862068965517243, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7724471830985915, + "global/total_cost": 0.0, + "num_tokens": 19186651.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.953125, + "simpleqa_verified/avg_subtask_length": 98.76283464566929, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.12146050670640834, + "2": 0.39418777943368105, + "3": 0.4821162444113264, + "4": 0.0014903129657228018, + "5": 0.0007451564828614009 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.046875, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.01921259842519685, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6286614173228346, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.037480314960629924, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.33385826771653543, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9940387481371088, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0007451564828614009, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.005216095380029807, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15521628498727735, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.030534351145038167, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8142493638676844, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7307692307692307, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.12615384615384614, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14307692307692307, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7045454545454546, + "simpleqa_verified/total_cost": 0.0, + "step": 70, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.8980978260869565, + "swe_bench/avg_subtask_length": 135.56833130328866, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0015128593040847202, + "2": 0.037821482602118005, + "3": 0.848714069591528, + "4": 0.07791225416036308, + "5": 0.0340393343419062 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.10190217391304347, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.02825822168087698, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30182704019488427, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5249695493300852, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17320341047503046, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2783661119515885, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.710287443267776, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.0113464447806354, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4106060606060606, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4863636363636364, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.10303030303030303, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2141732283464567, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.3732283464566929, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4125984251968504, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3310810810810811, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4594594594594595, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20945945945945946, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.17777777777777778, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7111111111111111, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.1111111111111111, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.7072010869565217, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.25309064239263535, + "epoch": 0.23202614379084968, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.7708140015602112, + "learning_rate": 7.546600811643816e-07, + "loss": 0.0133, + "reward": 1.1640625, + "reward_std": 0.12063967436552048, + "rewards/MultiTurnTaskReward/mean": 1.1640625, + "rewards/MultiTurnTaskReward/std": 0.47417303919792175, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0002710819244385, + "sampling/importance_sampling_ratio/min": 0.09943383187055588, + "sampling/sampling_logp_difference/max": 2.308262825012207, + "sampling/sampling_logp_difference/mean": 0.016523562371730804, + "step": 71 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9663461538461539, + "aime/avg_subtask_length": 146.4731954411144, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0024875621890547263, + "2": 0.10634328358208955, + "3": 0.8407960199004975, + "4": 0.04291044776119403, + "5": 0.007462686567164179 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03365384615384615, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.003588011819333052, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 16, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2728999577880962, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.01561840439003799, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7114816378218658, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.023009950248756218, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9769900497512438, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4096009975062344, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.00997506234413965, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.580423940149626, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.39706908583391487, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.034891835310537335, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5680390788555478, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.345679012345679, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.09876543209876543, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5555555555555556, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8876201923076923, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 438.0, + "completions/max_terminated_length": 438.0, + "completions/mean_length": 168.828125, + "completions/mean_terminated_length": 168.828125, + "completions/min_length": 66.0, + "completions/min_terminated_length": 66.0, + "epoch": 0.23202614379084968, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9409722222222222, + "global/avg_subtask_length": 130.0172032440403, + "global/chosen_num_routing_steps_ratio": { + "1": 0.0389760147601476, + "2": 0.17550738007380073, + "3": 0.7306273062730627, + "4": 0.041051660516605165, + "5": 0.013837638376383764 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.059027777777777776, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.015892520684852952, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 16, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.37781600720897845, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1977553862537888, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.42442860653723274, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.41051660516605165, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2220940959409594, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.36738929889298894, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.33405327573794097, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.17254619630429566, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4934005279577634, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3954785672342924, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18173810921902525, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4227833235466823, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3445378151260504, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3319327731092437, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3235294117647059, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.18333333333333332, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5666666666666667, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.25, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7756076388888888, + "global/total_cost": 0.0, + "num_tokens": 19419205.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9541666666666667, + "simpleqa_verified/avg_subtask_length": 98.6786261882858, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.11863173216885008, + "2": 0.3922852983988355, + "3": 0.4868995633187773, + "4": 0.001455604075691412, + "5": 0.000727802037845706 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.04583333333333333, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.01870591842992947, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6292548298068077, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.036491873658387, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3342532965348053, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9941775836972343, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.000727802037845706, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.005094614264919942, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15441783649876134, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.02972749793559042, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8158546655656482, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7366071428571429, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.12202380952380952, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14136904761904762, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7111111111111111, + "simpleqa_verified/total_cost": 0.0, + "step": 71, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.9002659574468085, + "swe_bench/avg_subtask_length": 135.7744771863118, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0014771048744460858, + "2": 0.03766617429837518, + "3": 0.8471196454948301, + "4": 0.07902511078286559, + "5": 0.03471196454948301 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.09973404255319149, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.027566539923954372, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.30109315589353614, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5278041825095057, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.17110266159695817, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.27843426883308714, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7104874446085672, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.011078286558345642, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.40532544378698226, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.4933431952662722, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.10133136094674557, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2175249807840123, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.37432744043043814, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4081475787855496, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.33116883116883117, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.461038961038961, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2077922077922078, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1702127659574468, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.723404255319149, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.10638297872340426, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.7134308510638298, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.22291807737201452, + "epoch": 0.23529411764705882, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.6276487112045288, + "learning_rate": 7.476589988449938e-07, + "loss": 0.0007, + "reward": 1.4921875, + "reward_std": 0.04419417306780815, + "rewards/MultiTurnTaskReward/mean": 1.4921875, + "rewards/MultiTurnTaskReward/std": 0.0883883461356163, + "sampling/importance_sampling_ratio/max": 1.8141772747039795, + "sampling/importance_sampling_ratio/mean": 0.9994735717773438, + "sampling/importance_sampling_ratio/min": 0.30038583278656006, + "sampling/sampling_logp_difference/max": 1.2026875019073486, + "sampling/sampling_logp_difference/mean": 0.014675293117761612, + "step": 72 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9669811320754716, + "aime/avg_subtask_length": 146.57412008281574, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0024390243902439024, + "2": 0.10670731707317073, + "3": 0.8414634146341463, + "4": 0.04207317073170732, + "5": 0.007317073170731708 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.0330188679245283, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0035196687370600416, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 17, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2732919254658385, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.015320910973084885, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7113871635610766, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.023170731707317073, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9768292682926829, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4077017114914425, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.009779951100244499, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5825183374083129, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4004106776180698, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.034223134839151265, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5653661875427789, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.345679012345679, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.09876543209876543, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5555555555555556, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8897405660377359, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 406.0, + "completions/max_terminated_length": 406.0, + "completions/mean_length": 188.4140625, + "completions/mean_terminated_length": 188.4140625, + "completions/min_length": 72.0, + "completions/min_terminated_length": 72.0, + "epoch": 0.23529411764705882, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9417808219178082, + "global/avg_subtask_length": 130.28502298572465, + "global/chosen_num_routing_steps_ratio": { + "1": 0.03840909090909091, + "2": 0.17386363636363636, + "3": 0.7329545454545454, + "4": 0.04090909090909091, + "5": 0.013863636363636364 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.05821917808219178, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.015646423098636987, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 17, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3757561093636584, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.20041938866037584, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.42382450197596583, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.40636363636363637, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22454545454545455, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3690909090909091, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.33183644528480266, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1767903568896242, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49137319782557315, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3958453548759377, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18320830929024812, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4209463358338142, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.34439834024896265, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.33195020746887965, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3236514522821577, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.19672131147540983, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5573770491803278, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.2459016393442623, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.778681506849315, + "global/total_cost": 0.0, + "num_tokens": 19696890.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9541666666666667, + "simpleqa_verified/avg_subtask_length": 98.6786261882858, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.11863173216885008, + "2": 0.3922852983988355, + "3": 0.4868995633187773, + "4": 0.001455604075691412, + "5": 0.000727802037845706 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.04583333333333333, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.01870591842992947, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6292548298068077, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.036491873658387, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3342532965348053, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9941775836972343, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.000727802037845706, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.005094614264919942, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15441783649876134, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.02972749793559042, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8158546655656482, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7366071428571429, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.12202380952380952, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14136904761904762, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7111111111111111, + "simpleqa_verified/total_cost": 0.0, + "step": 72, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.90234375, + "swe_bench/avg_subtask_length": 135.94707520891365, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.001443001443001443, + "2": 0.0367965367965368, + "3": 0.8484848484848485, + "4": 0.07864357864357864, + "5": 0.03463203463203463 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.09765625, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.02692664809656453, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.29874651810584957, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.532033426183844, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.1692200557103064, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.27705627705627706, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7121212121212122, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.010822510822510822, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.39739884393063585, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.5028901734104047, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.09971098265895954, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.21905476369092272, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.377344336084021, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4036009002250563, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.33121019108280253, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4585987261146497, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.21019108280254778, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1875, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7083333333333334, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.10416666666666667, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.7194010416666666, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.2804908836260438, + "epoch": 0.238562091503268, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.5150030851364136, + "learning_rate": 7.405929722454025e-07, + "loss": -0.0014, + "reward": 1.4453125, + "reward_std": 0.10500335693359375, + "rewards/MultiTurnTaskReward/mean": 1.4453125, + "rewards/MultiTurnTaskReward/std": 0.22826264798641205, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.000359296798706, + "sampling/importance_sampling_ratio/min": 0.4159519374370575, + "sampling/sampling_logp_difference/max": 1.3742709159851074, + "sampling/sampling_logp_difference/mean": 0.017518432810902596, + "step": 73 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9675925925925926, + "aime/avg_subtask_length": 147.57705583756345, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0023923444976076554, + "2": 0.10586124401913875, + "3": 0.8427033492822966, + "4": 0.041866028708133975, + "5": 0.007177033492822967 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.032407407407407406, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0034517766497461928, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 17, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.273502538071066, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.015025380710659899, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7114720812182741, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.022727272727272728, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9772727272727273, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.40707434052757796, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.009592326139088728, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5833333333333334, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4024144869215292, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.0335345405767941, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5640509725016767, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.34146341463414637, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0975609756097561, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5609756097560976, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8917824074074074, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 429.0, + "completions/max_terminated_length": 429.0, + "completions/mean_length": 217.6484375, + "completions/mean_terminated_length": 217.6484375, + "completions/min_length": 93.0, + "completions/min_terminated_length": 93.0, + "epoch": 0.238562091503268, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9425675675675675, + "global/avg_subtask_length": 130.99039225027792, + "global/chosen_num_routing_steps_ratio": { + "1": 0.037858422939068104, + "2": 0.17181899641577061, + "3": 0.7354390681003584, + "4": 0.04099462365591398, + "5": 0.013888888888888888 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.057432432432432436, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.015404160711449897, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 17, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3739082102588534, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.20303319040813086, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4230585993330157, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4014336917562724, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22759856630824374, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3709677419354839, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3310826542491269, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.18020954598370198, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4887077997671711, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3968253968253968, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18395691609977324, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4192176870748299, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.33877551020408164, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.33877551020408164, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3224489795918367, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1935483870967742, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5645161290322581, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.24193548387096775, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7816722972972973, + "global/total_cost": 0.0, + "num_tokens": 19986957.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9541666666666667, + "simpleqa_verified/avg_subtask_length": 98.6786261882858, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.11863173216885008, + "2": 0.3922852983988355, + "3": 0.4868995633187773, + "4": 0.001455604075691412, + "5": 0.000727802037845706 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.04583333333333333, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.01870591842992947, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6292548298068077, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.036491873658387, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3342532965348053, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9941775836972343, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.000727802037845706, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.005094614264919942, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15441783649876134, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.02972749793559042, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8158546655656482, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7366071428571429, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.12202380952380952, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14136904761904762, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.7111111111111111, + "simpleqa_verified/total_cost": 0.0, + "step": 73, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.9043367346938775, + "swe_bench/avg_subtask_length": 136.36229582577133, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0014104372355430183, + "2": 0.03596614950634697, + "3": 0.8497884344146686, + "4": 0.07827926657263752, + "5": 0.03455571227080395 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.09566326530612244, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.02631578947368421, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2971869328493648, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5362976406533575, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.16651542649727769, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2736248236953456, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7157968970380818, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.010578279266572637, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3926553672316384, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.5098870056497176, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.09745762711864407, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.22344322344322345, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.37875457875457874, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3978021978021978, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.325, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46875, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20625, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1836734693877551, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.10204081632653061, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.7251275510204082, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.2544110603630543, + "epoch": 0.24183006535947713, + "frac_reward_zero_std": 0.75, + "grad_norm": 1.2155472040176392, + "learning_rate": 7.334638543086203e-07, + "loss": -0.0058, + "reward": 1.4609375, + "reward_std": 0.09222550690174103, + "rewards/MultiTurnTaskReward/mean": 1.4609375, + "rewards/MultiTurnTaskReward/std": 0.194504976272583, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0003126859664917, + "sampling/importance_sampling_ratio/min": 0.31130504608154297, + "sampling/sampling_logp_difference/max": 1.1713404655456543, + "sampling/sampling_logp_difference/mean": 0.016197290271520615, + "step": 74 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9681818181818181, + "aime/avg_subtask_length": 147.33924656169026, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.002347417840375587, + "2": 0.10622065727699531, + "3": 0.8433098591549296, + "4": 0.04107981220657277, + "5": 0.007042253521126761 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.031818181818181815, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0033884791708192145, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 17, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.27366952361969304, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.014949172812437711, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7113813035678692, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.022300469483568074, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9776995305164319, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.40941176470588236, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.009411764705882352, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5811764705882353, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4009216589861751, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03357472021066491, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.56550362080316, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.34146341463414637, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0975609756097561, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5609756097560976, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.89375, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 424.0, + "completions/max_terminated_length": 424.0, + "completions/mean_length": 177.84375, + "completions/mean_terminated_length": 177.84375, + "completions/min_length": 82.0, + "completions/min_terminated_length": 82.0, + "epoch": 0.24183006535947713, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9433333333333334, + "global/avg_subtask_length": 130.7459658467805, + "global/chosen_num_routing_steps_ratio": { + "1": 0.03732332155477032, + "2": 0.17380742049469966, + "3": 0.7347614840989399, + "4": 0.040415194346289755, + "5": 0.013692579505300354 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.056666666666666664, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.015196616011279963, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 17, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3744320852263826, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.20045433181889394, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.42511358295472346, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4028268551236749, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22438162544169613, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.37279151943462896, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3305804083505391, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.1775636613902271, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49185593025923374, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.3975363941769317, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1822508398656215, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.42021276595744683, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.33877551020408164, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.33877551020408164, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3224489795918367, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1935483870967742, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5645161290322581, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.24193548387096775, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7845833333333333, + "global/total_cost": 0.0, + "num_tokens": 20220889.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9551630434782609, + "simpleqa_verified/avg_subtask_length": 98.41873690511822, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.11593172119487909, + "2": 0.39473684210526316, + "3": 0.4871977240398293, + "4": 0.001422475106685633, + "5": 0.0007112375533428165 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.04483695652173913, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.018258006584854835, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6276563903023047, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.03591739000299311, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3364262196947022, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9943100995732574, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0007112375533428165, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.004978662873399715, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15205148833467416, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.028962188254223652, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8189863234111022, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7354651162790697, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.12063953488372094, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.1438953488372093, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.717391304347826, + "simpleqa_verified/total_cost": 0.0, + "step": 74, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.9043367346938775, + "swe_bench/avg_subtask_length": 136.36229582577133, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0014104372355430183, + "2": 0.03596614950634697, + "3": 0.8497884344146686, + "4": 0.07827926657263752, + "5": 0.03455571227080395 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.09566326530612244, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.02631578947368421, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2971869328493648, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5362976406533575, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.16651542649727769, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2736248236953456, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7157968970380818, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.010578279266572637, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3926553672316384, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.5098870056497176, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.09745762711864407, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.22344322344322345, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.37875457875457874, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3978021978021978, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.325, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46875, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20625, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1836734693877551, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.10204081632653061, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.7251275510204082, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.27060729544609785, + "epoch": 0.24509803921568626, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.854928195476532, + "learning_rate": 7.262735145222695e-07, + "loss": -0.0359, + "reward": 1.4765625, + "reward_std": 0.07403614372015, + "rewards/MultiTurnTaskReward/mean": 1.4765625, + "rewards/MultiTurnTaskReward/std": 0.15188287198543549, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0002243518829346, + "sampling/importance_sampling_ratio/min": 0.17685134708881378, + "sampling/sampling_logp_difference/max": 1.7324457168579102, + "sampling/sampling_logp_difference/mean": 0.0157189778983593, + "step": 75 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.96875, + "aime/avg_subtask_length": 147.68270735524257, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.002304147465437788, + "2": 0.1054147465437788, + "3": 0.8444700460829493, + "4": 0.04089861751152074, + "5": 0.0069124423963133645 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03125, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.003325508607198748, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 17, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2734741784037559, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.014671361502347418, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7118544600938967, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.021889400921658985, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.978110599078341, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4053117782909931, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.009237875288683603, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5854503464203233, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4054228534538412, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.0329244673983215, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5616526791478373, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3373493975903614, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0963855421686747, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5662650602409639, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8956473214285714, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 241.0, + "completions/max_terminated_length": 241.0, + "completions/mean_length": 145.8046875, + "completions/mean_terminated_length": 145.8046875, + "completions/min_length": 82.0, + "completions/min_terminated_length": 82.0, + "epoch": 0.24509803921568626, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.944078947368421, + "global/avg_subtask_length": 130.78342865976194, + "global/chosen_num_routing_steps_ratio": { + "1": 0.036803135888501745, + "2": 0.1759581881533101, + "3": 0.7336672473867596, + "4": 0.04006968641114982, + "5": 0.013501742160278746 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.05592105263157895, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.014994589581078991, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 17, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.37509661462358945, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.1977894574122739, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.42711392796413666, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.40418118466898956, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22125435540069685, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.37456445993031356, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3285100610445399, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.17499434772778658, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4964955912276735, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4008298755186722, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18008298755186722, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4190871369294606, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.33739837398373984, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.33739837398373984, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3252032520325203, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1935483870967742, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5645161290322581, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.24193548387096775, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7874177631578947, + "global/total_cost": 0.0, + "num_tokens": 20416640.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.9561170212765957, + "simpleqa_verified/avg_subtask_length": 98.31392627267408, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.1133518776077886, + "2": 0.3991655076495132, + "3": 0.4853963838664812, + "4": 0.0013908205841446453, + "5": 0.0006954102920723226 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.043882978723404256, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.017846693973083673, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6275599765944997, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.03510825043885313, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3373317729666472, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9944367176634215, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0006954102920723226, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.004867872044506259, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15294117647058825, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.02823529411764706, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8188235294117647, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7360912981455064, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.11840228245363767, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14550641940085593, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.723404255319149, + "simpleqa_verified/total_cost": 0.0, + "step": 75, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.9043367346938775, + "swe_bench/avg_subtask_length": 136.36229582577133, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.0014104372355430183, + "2": 0.03596614950634697, + "3": 0.8497884344146686, + "4": 0.07827926657263752, + "5": 0.03455571227080395 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.09566326530612244, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.02631578947368421, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2971869328493648, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5362976406533575, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.16651542649727769, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2736248236953456, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7157968970380818, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.010578279266572637, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3926553672316384, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.5098870056497176, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.09745762711864407, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.22344322344322345, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.37875457875457874, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3978021978021978, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.325, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.46875, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20625, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1836734693877551, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.10204081632653061, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.7251275510204082, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.27660490199923515, + "epoch": 0.24836601307189543, + "frac_reward_zero_std": 1.0, + "grad_norm": 0.0, + "learning_rate": 7.190238384283412e-07, + "loss": 0.0, + "reward": 1.5, + "reward_std": 0.0, + "rewards/MultiTurnTaskReward/mean": 1.5, + "rewards/MultiTurnTaskReward/std": 0.0, + "sampling/importance_sampling_ratio/max": 1.5830142498016357, + "sampling/importance_sampling_ratio/mean": 0.9996315240859985, + "sampling/importance_sampling_ratio/min": 0.5678392052650452, + "sampling/sampling_logp_difference/max": 0.5659170150756836, + "sampling/sampling_logp_difference/mean": 0.015026800334453583, + "step": 76 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.96875, + "aime/avg_subtask_length": 147.68270735524257, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.002304147465437788, + "2": 0.1054147465437788, + "3": 0.8444700460829493, + "4": 0.04089861751152074, + "5": 0.0069124423963133645 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03125, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.003325508607198748, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 17, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2734741784037559, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.014671361502347418, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7118544600938967, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.021889400921658985, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.978110599078341, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4053117782909931, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.009237875288683603, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5854503464203233, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4054228534538412, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.0329244673983215, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5616526791478373, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3373493975903614, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0963855421686747, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5662650602409639, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8956473214285714, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 353.0, + "completions/max_terminated_length": 353.0, + "completions/mean_length": 180.046875, + "completions/mean_terminated_length": 180.046875, + "completions/min_length": 84.0, + "completions/min_terminated_length": 84.0, + "epoch": 0.24836601307189543, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9448051948051948, + "global/avg_subtask_length": 130.70823798627003, + "global/chosen_num_routing_steps_ratio": { + "1": 0.03629725085910653, + "2": 0.17804982817869416, + "3": 0.7326030927835051, + "4": 0.039733676975945015, + "5": 0.01331615120274914 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.05519480519480519, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.014797864225781845, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 17, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.37513348588863465, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.20030511060259343, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4245614035087719, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4065721649484536, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22401202749140894, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3694158075601375, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3264987742366838, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.17829284599955428, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.495208379763762, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.400492072170585, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1817933296883543, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.4177145981410607, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3360323886639676, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3360323886639676, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.32793522267206476, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1935483870967742, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5645161290322581, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.24193548387096775, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7834821428571429, + "global/total_cost": 0.0, + "num_tokens": 20653766.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.95703125, + "simpleqa_verified/avg_subtask_length": 98.38963641568851, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.1108843537414966, + "2": 0.40476190476190477, + "3": 0.482312925170068, + "4": 0.0013605442176870747, + "5": 0.0006802721088435374 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.04296875, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.017463498425422275, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.627254509018036, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.03435442313197824, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.3383910678499857, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9945578231292517, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0006802721088435374, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.004761904761904762, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.1545524100994644, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.02754399387911247, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8179035960214232, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7345505617977528, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.11657303370786516, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.14887640449438203, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.708984375, + "simpleqa_verified/total_cost": 0.0, + "step": 76, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.90625, + "swe_bench/avg_subtask_length": 136.50521642619313, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.001379310344827586, + "2": 0.03517241379310345, + "3": 0.8524137931034482, + "4": 0.07724137931034483, + "5": 0.03379310344827586 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.09375, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.02574916759156493, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.29500554938956713, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.539622641509434, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.16537180910099888, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2710344827586207, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7186206896551725, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.010344827586206896, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.38743093922651933, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.5165745856353591, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.09599447513812155, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.22476735862562633, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.380100214745884, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.39513242662848963, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.32298136645962733, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4658385093167702, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2111801242236025, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1836734693877551, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.10204081632653061, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.729375, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.2761524999514222, + "epoch": 0.25163398692810457, + "frac_reward_zero_std": 0.25, + "grad_norm": 1.4465316534042358, + "learning_rate": 7.117167271287452e-07, + "loss": -0.008, + "reward": 1.09375, + "reward_std": 0.23042556643486023, + "rewards/MultiTurnTaskReward/mean": 1.09375, + "rewards/MultiTurnTaskReward/std": 0.4930621087551117, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9997127056121826, + "sampling/importance_sampling_ratio/min": 0.3785257935523987, + "sampling/sampling_logp_difference/max": 0.9714710712432861, + "sampling/sampling_logp_difference/mean": 0.015617300756275654, + "step": 77 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.96875, + "aime/avg_subtask_length": 147.68270735524257, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.002304147465437788, + "2": 0.1054147465437788, + "3": 0.8444700460829493, + "4": 0.04089861751152074, + "5": 0.0069124423963133645 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03125, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.003325508607198748, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 17, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2734741784037559, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.014671361502347418, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7118544600938967, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.021889400921658985, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.978110599078341, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4053117782909931, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.009237875288683603, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5854503464203233, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4054228534538412, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.0329244673983215, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5616526791478373, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3373493975903614, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0963855421686747, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5662650602409639, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8956473214285714, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 342.0, + "completions/max_terminated_length": 342.0, + "completions/mean_length": 169.1953125, + "completions/mean_terminated_length": 169.1953125, + "completions/min_length": 83.0, + "completions/min_terminated_length": 83.0, + "epoch": 0.25163398692810457, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9455128205128205, + "global/avg_subtask_length": 130.46840400692926, + "global/chosen_num_routing_steps_ratio": { + "1": 0.03580508474576271, + "2": 0.18093220338983051, + "3": 0.7309322033898306, + "4": 0.03919491525423729, + "5": 0.013135593220338982 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.05448717948717949, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.014611734578594562, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 17, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.3778715071175717, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.19778564434736762, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.42434284853506066, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.41461864406779664, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.22097457627118644, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3644067796610169, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3232256646890793, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.17578554163920018, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5009887936717206, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4041114417094942, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.17987557479037056, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41601298350013527, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3360323886639676, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.3360323886639676, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.32793522267206476, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1935483870967742, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5645161290322581, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.24193548387096775, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7848557692307693, + "global/total_cost": 0.0, + "num_tokens": 20890335.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.95875, + "simpleqa_verified/avg_subtask_length": 98.99426229508197, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.10625814863102999, + "2": 0.4041720990873533, + "3": 0.4876140808344198, + "4": 0.001303780964797914, + "5": 0.000651890482398957 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.04125, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.016666666666666666, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6256830601092896, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.03278688524590164, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.34153005464480873, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9947848761408083, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.000651890482398957, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.0045632333767926985, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15171407731582787, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0262582056892779, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8220277169948942, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7350199733688415, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.11051930758988016, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.15446071904127828, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.71625, + "simpleqa_verified/total_cost": 0.0, + "step": 77, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.90625, + "swe_bench/avg_subtask_length": 136.50521642619313, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.001379310344827586, + "2": 0.03517241379310345, + "3": 0.8524137931034482, + "4": 0.07724137931034483, + "5": 0.03379310344827586 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.09375, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.02574916759156493, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.29500554938956713, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.539622641509434, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.16537180910099888, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2710344827586207, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7186206896551725, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.010344827586206896, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.38743093922651933, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.5165745856353591, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.09599447513812155, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.22476735862562633, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.380100214745884, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.39513242662848963, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.32298136645962733, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4658385093167702, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2111801242236025, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1836734693877551, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.10204081632653061, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.729375, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.2077170042321086, + "epoch": 0.2549019607843137, + "frac_reward_zero_std": 0.5, + "grad_norm": 0.818428635597229, + "learning_rate": 7.043540967867781e-07, + "loss": -0.0105, + "reward": 1.4375, + "reward_std": 0.1491975337266922, + "rewards/MultiTurnTaskReward/mean": 1.4375, + "rewards/MultiTurnTaskReward/std": 0.24301259219646454, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 1.0000149011611938, + "sampling/importance_sampling_ratio/min": 0.5038020610809326, + "sampling/sampling_logp_difference/max": 0.9791951179504395, + "sampling/sampling_logp_difference/mean": 0.013623332604765892, + "step": 78 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.96875, + "aime/avg_subtask_length": 147.68270735524257, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.002304147465437788, + "2": 0.1054147465437788, + "3": 0.8444700460829493, + "4": 0.04089861751152074, + "5": 0.0069124423963133645 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03125, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.003325508607198748, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 17, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2734741784037559, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.014671361502347418, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7118544600938967, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.021889400921658985, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.978110599078341, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.4053117782909931, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.009237875288683603, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5854503464203233, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.4054228534538412, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.0329244673983215, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5616526791478373, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3373493975903614, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0963855421686747, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5662650602409639, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8956473214285714, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 466.0, + "completions/max_terminated_length": 466.0, + "completions/mean_length": 239.9140625, + "completions/mean_terminated_length": 239.9140625, + "completions/min_length": 91.0, + "completions/min_terminated_length": 91.0, + "epoch": 0.2549019607843137, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.9462025316455697, + "global/avg_subtask_length": 130.62961038961038, + "global/chosen_num_routing_steps_ratio": { + "1": 0.035326086956521736, + "2": 0.17851170568561872, + "3": 0.7332775919732442, + "4": 0.0399247491638796, + "5": 0.012959866220735786 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.05379746835443038, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.014397031539888683, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 17, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.37551020408163266, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.20534322820037107, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4191465677179963, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.41095317725752506, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2295150501672241, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3595317725752508, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3228602383531961, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.18309859154929578, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.49404117009750814, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.40095719223610743, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.1863865993086945, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41265620845519807, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.33201581027667987, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.34782608695652173, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.3201581027667984, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1935483870967742, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5645161290322581, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.24193548387096775, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7784810126582279, + "global/total_cost": 0.0, + "num_tokens": 21248692.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.95875, + "simpleqa_verified/avg_subtask_length": 98.99426229508197, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.10625814863102999, + "2": 0.4041720990873533, + "3": 0.4876140808344198, + "4": 0.001303780964797914, + "5": 0.000651890482398957 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.04125, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.016666666666666666, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6256830601092896, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.03278688524590164, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.34153005464480873, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9947848761408083, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.000651890482398957, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.0045632333767926985, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15171407731582787, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0262582056892779, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8220277169948942, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7350199733688415, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.11051930758988016, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.15446071904127828, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.71625, + "simpleqa_verified/total_cost": 0.0, + "step": 78, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.9098557692307693, + "swe_bench/avg_subtask_length": 136.7129491813736, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.001321003963011889, + "2": 0.03368560105680317, + "3": 0.8546895640686922, + "4": 0.07793923381770146, + "5": 0.03236459709379128 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.09014423076923077, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.02466510737826919, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.29172868381883904, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5468849670423134, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.16138634913884756, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2655217965653897, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.7245706737120211, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.009907529722589167, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3835978835978836, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.5244708994708994, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.09193121693121693, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.2245037645448323, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.38809034907597534, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.3874058863791923, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.31736526946107785, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.47904191616766467, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.20359281437125748, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1836734693877551, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.10204081632653061, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.7121394230769231, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.18847736902534962, + "epoch": 0.2581699346405229, + "frac_reward_zero_std": 0.75, + "grad_norm": 0.6140043139457703, + "learning_rate": 6.969378781246436e-07, + "loss": 0.0162, + "reward": 0.890625, + "reward_std": 0.12600402534008026, + "rewards/MultiTurnTaskReward/mean": 0.890625, + "rewards/MultiTurnTaskReward/std": 0.4898075461387634, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9994298219680786, + "sampling/importance_sampling_ratio/min": 0.2949640154838562, + "sampling/sampling_logp_difference/max": 1.350264549255371, + "sampling/sampling_logp_difference/mean": 0.013667132705450058, + "step": 79 + }, + { + "aime/all_access_ratio": 1.0, + "aime/avg_format_bonus": 0.9692982456140351, + "aime/avg_subtask_length": 148.02151363810987, + "aime/chosen_num_routing_steps_ratio": { + "1": 0.0022624434389140274, + "2": 0.10463800904977376, + "3": 0.8461538461538461, + "4": 0.040158371040723985, + "5": 0.006787330316742082 + }, + "aime/complex_topology_future_error_cnt": 0.0, + "aime/format_error_ratio": 0.03070175438596491, + "aime/interrupt_ratio": 0.0, + "aime/invalid_access_ratio": 0.0032654629273914716, + "aime/invalid_ctop_access_cnt": 0.0, + "aime/model_id_error_ratio": 0.0, + "aime/non_chain_topology_ratio": 0.0, + "aime/none_returns/gemini/gemini-3-pro-preview": 6, + "aime/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "aime/none_returns/openrouter/openai/gpt-5.2": 17, + "aime/request_timeout_cnt": 0, + "aime/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2748751440645409, + "aime/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.014406454091432963, + "aime/router_selection_ratio/openrouter/openai/gpt-5.2": 0.7107184018440261, + "aime/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.022058823529411766, + "aime/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.9779411764705882, + "aime/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.41099773242630383, + "aime/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.009070294784580499, + "aime/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.5799319727891157, + "aime/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.40341988600379985, + "aime/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.03229892336922103, + "aime/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.5642811906269791, + "aime/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3373493975903614, + "aime/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0963855421686747, + "aime/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.5662650602409639, + "aime/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.16666666666666666, + "aime/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.8333333333333334, + "aime/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "aime/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "aime/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "aime/router_selection_visibility_ratio_pos1": 0.0, + "aime/router_selection_visibility_ratio_pos2": 0.0, + "aime/router_selection_visibility_ratio_pos3": 0.0, + "aime/router_selection_visibility_ratio_pos4": 0.0, + "aime/router_selection_visibility_ratio_pos5": 0.0, + "aime/subtask_misspecification_ratio": 0.0, + "aime/total_correct_ratio": 0.8974780701754386, + "aime/total_cost": 0.0, + "completions/clipped_ratio": 0.0, + "completions/max_length": 416.0, + "completions/max_terminated_length": 416.0, + "completions/mean_length": 214.5859375, + "completions/mean_terminated_length": 214.5859375, + "completions/min_length": 100.0, + "completions/min_terminated_length": 100.0, + "epoch": 0.2581699346405229, + "global/all_access_ratio": 1.0, + "global/avg_format_bonus": 0.946875, + "global/avg_subtask_length": 131.0735294117647, + "global/chosen_num_routing_steps_ratio": { + "1": 0.03485973597359736, + "2": 0.17656765676567657, + "3": 0.7357673267326733, + "4": 0.040016501650165015, + "5": 0.01278877887788779 + }, + "global/complex_topology_future_error_cnt": 0.0, + "global/format_error_ratio": 0.053125, + "global/interrupt_ratio": 0.0, + "global/invalid_access_ratio": 0.014193737196371086, + "global/invalid_ctop_access_cnt": 0.0, + "global/model_id_error_ratio": 0.0, + "global/non_chain_topology_ratio": 0.0, + "global/none_returns/gemini/gemini-3-pro-preview": 6, + "global/none_returns/openrouter/anthropic/claude-opus-4.5": 1, + "global/none_returns/openrouter/openai/gpt-5.2": 17, + "global/request_timeout_cnt": 0, + "global/router_selection_ratio/gemini/gemini-3-pro-preview": 0.37452443664032775, + "global/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.20712613403570385, + "global/router_selection_ratio/openrouter/openai/gpt-5.2": 0.4183494293239684, + "global/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.4082095709570957, + "global/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.2306105610561056, + "global/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.3611798679867987, + "global/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.3240008548835221, + "global/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.18679204958324427, + "global/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.4892070955332336, + "global/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.39916296102537274, + "global/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.18702589589327753, + "global/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.41381114308134975, + "global/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3359375, + "global/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.34765625, + "global/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.31640625, + "global/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1935483870967742, + "global/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.5645161290322581, + "global/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.24193548387096775, + "global/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "global/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "global/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "global/router_selection_visibility_ratio_pos1": 0.0, + "global/router_selection_visibility_ratio_pos2": 0.0, + "global/router_selection_visibility_ratio_pos3": 0.0, + "global/router_selection_visibility_ratio_pos4": 0.0, + "global/router_selection_visibility_ratio_pos5": 0.0, + "global/subtask_misspecification_ratio": 0.0, + "global/total_correct_ratio": 0.7751953125, + "global/total_cost": 0.0, + "num_tokens": 21564191.0, + "simpleqa_verified/all_access_ratio": 1.0, + "simpleqa_verified/avg_format_bonus": 0.95875, + "simpleqa_verified/avg_subtask_length": 98.99426229508197, + "simpleqa_verified/chosen_num_routing_steps_ratio": { + "1": 0.10625814863102999, + "2": 0.4041720990873533, + "3": 0.4876140808344198, + "4": 0.001303780964797914, + "5": 0.000651890482398957 + }, + "simpleqa_verified/complex_topology_future_error_cnt": 0.0, + "simpleqa_verified/format_error_ratio": 0.04125, + "simpleqa_verified/interrupt_ratio": 0.0, + "simpleqa_verified/invalid_access_ratio": 0.016666666666666666, + "simpleqa_verified/invalid_ctop_access_cnt": 0.0, + "simpleqa_verified/model_id_error_ratio": 0.0, + "simpleqa_verified/non_chain_topology_ratio": 0.0, + "simpleqa_verified/none_returns/gemini/gemini-3-pro-preview": 0, + "simpleqa_verified/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "simpleqa_verified/none_returns/openrouter/openai/gpt-5.2": 0, + "simpleqa_verified/request_timeout_cnt": 0, + "simpleqa_verified/router_selection_ratio/gemini/gemini-3-pro-preview": 0.6256830601092896, + "simpleqa_verified/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.03278688524590164, + "simpleqa_verified/router_selection_ratio/openrouter/openai/gpt-5.2": 0.34153005464480873, + "simpleqa_verified/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.9947848761408083, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.000651890482398957, + "simpleqa_verified/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.0045632333767926985, + "simpleqa_verified/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.15171407731582787, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.0262582056892779, + "simpleqa_verified/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.8220277169948942, + "simpleqa_verified/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.7350199733688415, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.11051930758988016, + "simpleqa_verified/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.15446071904127828, + "simpleqa_verified/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 1.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "simpleqa_verified/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos1": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos2": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos3": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos4": 0.0, + "simpleqa_verified/router_selection_visibility_ratio_pos5": 0.0, + "simpleqa_verified/subtask_misspecification_ratio": 0.0, + "simpleqa_verified/total_correct_ratio": 0.71625, + "simpleqa_verified/total_cost": 0.0, + "step": 79, + "swe_bench/all_access_ratio": 1.0, + "swe_bench/avg_format_bonus": 0.9115566037735849, + "swe_bench/avg_subtask_length": 137.1499375260308, + "swe_bench/chosen_num_routing_steps_ratio": { + "1": 0.00129366106080207, + "2": 0.032988357050452784, + "3": 0.8557567917205692, + "4": 0.07826649417852523, + "5": 0.03169469598965071 + }, + "swe_bench/complex_topology_future_error_cnt": 0.0, + "swe_bench/format_error_ratio": 0.08844339622641509, + "swe_bench/interrupt_ratio": 0.0, + "swe_bench/invalid_access_ratio": 0.024156601416076635, + "swe_bench/invalid_ctop_access_cnt": 0.0, + "swe_bench/model_id_error_ratio": 0.0, + "swe_bench/non_chain_topology_ratio": 0.0, + "swe_bench/none_returns/gemini/gemini-3-pro-preview": 0, + "swe_bench/none_returns/openrouter/anthropic/claude-opus-4.5": 0, + "swe_bench/none_returns/openrouter/openai/gpt-5.2": 0, + "swe_bench/request_timeout_cnt": 0, + "swe_bench/router_selection_ratio/gemini/gemini-3-pro-preview": 0.2911286963765098, + "swe_bench/router_selection_ratio/openrouter/anthropic/claude-opus-4.5": 0.5489379425239483, + "swe_bench/router_selection_ratio/openrouter/openai/gpt-5.2": 0.15993336109954187, + "swe_bench/router_selection_ratio_pos1/gemini/gemini-3-pro-preview": 0.2677878395860285, + "swe_bench/router_selection_ratio_pos1/openrouter/anthropic/claude-opus-4.5": 0.722509702457956, + "swe_bench/router_selection_ratio_pos1/openrouter/openai/gpt-5.2": 0.009702457956015523, + "swe_bench/router_selection_ratio_pos2/gemini/gemini-3-pro-preview": 0.37759067357512954, + "swe_bench/router_selection_ratio_pos2/openrouter/anthropic/claude-opus-4.5": 0.5323834196891192, + "swe_bench/router_selection_ratio_pos2/openrouter/openai/gpt-5.2": 0.0900259067357513, + "swe_bench/router_selection_ratio_pos3/gemini/gemini-3-pro-preview": 0.22572002679169456, + "swe_bench/router_selection_ratio_pos3/openrouter/anthropic/claude-opus-4.5": 0.38914936369725384, + "swe_bench/router_selection_ratio_pos3/openrouter/openai/gpt-5.2": 0.38513060951105155, + "swe_bench/router_selection_ratio_pos4/gemini/gemini-3-pro-preview": 0.3235294117647059, + "swe_bench/router_selection_ratio_pos4/openrouter/anthropic/claude-opus-4.5": 0.4764705882352941, + "swe_bench/router_selection_ratio_pos4/openrouter/openai/gpt-5.2": 0.2, + "swe_bench/router_selection_ratio_pos5/gemini/gemini-3-pro-preview": 0.1836734693877551, + "swe_bench/router_selection_ratio_pos5/openrouter/anthropic/claude-opus-4.5": 0.7142857142857143, + "swe_bench/router_selection_ratio_pos5/openrouter/openai/gpt-5.2": 0.10204081632653061, + "swe_bench/router_selection_visibility_ratio/gemini/gemini-3-pro-preview": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/anthropic/claude-opus-4.5": 0.0, + "swe_bench/router_selection_visibility_ratio/openrouter/openai/gpt-5.2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos1": 0.0, + "swe_bench/router_selection_visibility_ratio_pos2": 0.0, + "swe_bench/router_selection_visibility_ratio_pos3": 0.0, + "swe_bench/router_selection_visibility_ratio_pos4": 0.0, + "swe_bench/router_selection_visibility_ratio_pos5": 0.0, + "swe_bench/subtask_misspecification_ratio": 0.0, + "swe_bench/total_correct_ratio": 0.6992924528301887, + "swe_bench/total_cost": 0.0 + }, + { + "clip_ratio/high_max": 0.0, + "clip_ratio/high_mean": 0.0, + "clip_ratio/low_mean": 0.0, + "clip_ratio/low_min": 0.0, + "clip_ratio/region_mean": 0.0, + "entropy": 0.2456964459270239, + "epoch": 0.26143790849673204, + "frac_reward_zero_std": 0.5, + "grad_norm": 0.8849470019340515, + "learning_rate": 6.894700159171534e-07, + "loss": 0.0001, + "reward": 1.1875, + "reward_std": 0.15839502215385437, + "rewards/MultiTurnTaskReward/mean": 1.1875, + "rewards/MultiTurnTaskReward/std": 0.4653336703777313, + "sampling/importance_sampling_ratio/max": 2.0, + "sampling/importance_sampling_ratio/mean": 0.9999550580978394, + "sampling/importance_sampling_ratio/min": 0.3906385898590088, + "sampling/sampling_logp_difference/max": 0.9399725198745728, + "sampling/sampling_logp_difference/mean": 0.015304848551750183, + "step": 80 + } + ], + "logging_steps": 1, + "max_steps": 200, + "num_input_tokens_seen": 21564191, + "num_train_epochs": 1, + "save_steps": 20, + "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": 4, + "trial_name": null, + "trial_params": null +}