{ "aime2026": { "acc_len": 2.6565662068447096, "acc_len_max": 4.318471337579618, "acc_len_median": 2.6269313291099063, "acc_len_min": 1.9606270943034945, "concurrency": 1, "elapsed_s": 3335.8, "errors": 0, "max_tokens": 16384, "mean_completion_tokens": 6000, "mean_prompt_tokens": 207, "n_acc_matched": 30, "n_acc_unmatched": 0, "n_prompts": 30, "n_prompts_expected": 30, "n_requests": 30, "output_tokens": 179990, "output_tokens_per_s": 53.957, "prompt_policy": "raw dataset text, no instruction wrapper", "proposal_acceptance": 0.23665231526352995, "proposal_count": 7, "reasoning": "max", "row": "aime2026", "sampling": { "temperature": 1.0, "top_p": 0.95 }, "semantic_correct": 23, "semantic_graded": 30, "semantic_score": 0.7666666666666667, "truncated": 6 }, "gsm8k": { "acc_len": 5.851219913596805, "acc_len_max": 8.0, "acc_len_median": 5.913043478260869, "acc_len_min": 2.851063829787234, "concurrency": 1, "elapsed_s": 1535.7, "errors": 0, "max_tokens": 4096, "mean_completion_tokens": 247, "mean_prompt_tokens": 147, "n_acc_matched": 1319, "n_acc_unmatched": 0, "n_prompts": 1319, "n_prompts_expected": 1319, "n_requests": 1319, "output_tokens": 326026, "output_tokens_per_s": 212.296, "prompt_policy": "raw dataset text, no instruction wrapper", "proposal_acceptance": 0.6930314162281149, "proposal_count": 7, "reasoning": "max", "row": "gsm8k", "sampling": { "temperature": 1.0, "top_p": 0.95 }, "semantic_correct": 976, "semantic_graded": 1319, "semantic_score": 0.7399545109931767, "truncated": 0 }, "humaneval": { "acc_len": 4.866431561653052, "acc_len_max": 6.56989247311828, "acc_len_median": 4.980739360049705, "acc_len_min": 2.6048192771084335, "concurrency": 1, "elapsed_s": 810.0, "errors": 0, "max_tokens": 4096, "mean_completion_tokens": 999, "mean_prompt_tokens": 220, "n_acc_matched": 164, "n_acc_unmatched": 0, "n_prompts": 164, "n_prompts_expected": 164, "n_requests": 164, "output_tokens": 163891, "output_tokens_per_s": 202.325, "prompt_policy": "raw dataset text, no instruction wrapper", "proposal_acceptance": 0.552347365950436, "proposal_count": 7, "reasoning": "max", "row": "humaneval", "sampling": { "temperature": 1.0, "top_p": 0.95 }, "truncated": 1 }, "math500": { "acc_len": 4.571222086925199, "acc_len_max": 7.366666666666666, "acc_len_median": 4.629166666666666, "acc_len_min": 2.1540983606557376, "concurrency": 1, "elapsed_s": 3128.5, "errors": 0, "max_tokens": 8192, "mean_completion_tokens": 791, "mean_prompt_tokens": 156, "n_acc_matched": 500, "n_acc_unmatched": 0, "n_prompts": 500, "n_prompts_expected": 500, "n_requests": 500, "output_tokens": 395616, "output_tokens_per_s": 126.456, "prompt_policy": "raw dataset text, no instruction wrapper", "proposal_acceptance": 0.510174583846457, "proposal_count": 7, "reasoning": "max", "row": "math500", "sampling": { "temperature": 1.0, "top_p": 0.95 }, "semantic_correct": 336, "semantic_graded": 500, "semantic_score": 0.672, "truncated": 4 }, "mtbench": { "acc_len": 3.4807646201963163, "acc_len_max": 6.45945945945946, "acc_len_median": 3.0978285971726898, "acc_len_min": 2.0985915492957745, "concurrency": 1, "elapsed_s": 3010.8, "errors": 0, "max_tokens": 4096, "mean_completion_tokens": 1789, "mean_prompt_tokens": 944, "n_acc_matched": 160, "n_acc_unmatched": 0, "n_prompts": 80, "n_prompts_expected": 80, "n_requests": 160, "output_tokens": 286305, "output_tokens_per_s": 95.094, "prompt_policy": "raw dataset text, no instruction wrapper", "proposal_acceptance": 0.3543949457423309, "proposal_count": 7, "reasoning": "max", "row": "mtbench", "sampling": { "temperature": 1.0, "top_p": 0.95 }, "truncated": 21 }, "sb_coding": { "acc_len": 4.262703790885098, "acc_len_max": 5.730769230769231, "acc_len_median": 4.396610169491526, "acc_len_min": 2.6, "concurrency": 1, "elapsed_s": 774.4, "errors": 0, "max_tokens": 4096, "mean_completion_tokens": 1405, "mean_prompt_tokens": 405, "n_acc_matched": 89, "n_acc_unmatched": 0, "n_prompts": 80, "n_prompts_expected": 80, "n_requests": 89, "output_tokens": 125028, "output_tokens_per_s": 161.452, "prompt_policy": "raw dataset text, no instruction wrapper", "proposal_acceptance": 0.466100541555014, "proposal_count": 7, "reasoning": "max", "row": "sb_coding", "sampling": { "temperature": 1.0, "top_p": 0.95 }, "truncated": 3 }, "sb_multilingual": { "acc_len": 3.559084746910859, "acc_len_max": 5.517857142857143, "acc_len_median": 3.340179316096747, "acc_len_min": 2.4008875739644973, "concurrency": 1, "elapsed_s": 297.4, "errors": 0, "max_tokens": 4096, "mean_completion_tokens": 586, "mean_prompt_tokens": 179, "n_acc_matched": 80, "n_acc_unmatched": 0, "n_prompts": 80, "n_prompts_expected": 80, "n_requests": 80, "output_tokens": 46892, "output_tokens_per_s": 157.65, "prompt_policy": "raw dataset text, no instruction wrapper", "proposal_acceptance": 0.3655835352729798, "proposal_count": 7, "reasoning": "max", "row": "sb_multilingual", "sampling": { "temperature": 1.0, "top_p": 0.95 }, "truncated": 0 }, "sb_qa": { "acc_len": 3.1203881130019955, "acc_len_max": 4.268456375838926, "acc_len_median": 3.0877691387559807, "acc_len_min": 2.2061971830985914, "concurrency": 1, "elapsed_s": 597.6, "errors": 0, "max_tokens": 4096, "mean_completion_tokens": 985, "mean_prompt_tokens": 99, "n_acc_matched": 80, "n_acc_unmatched": 0, "n_prompts": 80, "n_prompts_expected": 80, "n_requests": 80, "output_tokens": 78817, "output_tokens_per_s": 131.898, "prompt_policy": "raw dataset text, no instruction wrapper", "proposal_acceptance": 0.30291258757171363, "proposal_count": 7, "reasoning": "max", "row": "sb_qa", "sampling": { "temperature": 1.0, "top_p": 0.95 }, "truncated": 2 }, "sb_rag": { "acc_len": 3.8188619710184173, "acc_len_max": 5.541666666666667, "acc_len_median": 3.7484909456740443, "acc_len_min": 2.709433962264151, "concurrency": 1, "elapsed_s": 427.5, "errors": 0, "max_tokens": 4096, "mean_completion_tokens": 650, "mean_prompt_tokens": 703, "n_acc_matched": 91, "n_acc_unmatched": 0, "n_prompts": 80, "n_prompts_expected": 80, "n_requests": 91, "output_tokens": 59139, "output_tokens_per_s": 138.34, "prompt_policy": "raw dataset text, no instruction wrapper", "proposal_acceptance": 0.40269456728834535, "proposal_count": 7, "reasoning": "max", "row": "sb_rag", "sampling": { "temperature": 1.0, "top_p": 0.95 }, "truncated": 1 }, "sb_writing": { "acc_len": 2.676241748033513, "acc_len_max": 4.743975903614458, "acc_len_median": 2.5655032035301577, "acc_len_min": 1.941678520625889, "concurrency": 1, "elapsed_s": 3406.6, "errors": 0, "max_tokens": 4096, "mean_completion_tokens": 3543, "mean_prompt_tokens": 785, "n_acc_matched": 84, "n_acc_unmatched": 0, "n_prompts": 80, "n_prompts_expected": 80, "n_requests": 84, "output_tokens": 297601, "output_tokens_per_s": 87.359, "prompt_policy": "raw dataset text, no instruction wrapper", "proposal_acceptance": 0.23946310686193042, "proposal_count": 7, "reasoning": "max", "row": "sb_writing", "sampling": { "temperature": 1.0, "top_p": 0.95 }, "truncated": 54 } }