{ "problems": 256, "samples": 512, "acc": 0.591796875, "acc_se": 0.02824871305347183, "cut_share": 0.291015625, "mean_tokens": 17729.541015625, "closed_share": 0.8359375, "p_finish": 0.708984375, "p_correct_given_finished": 0.8292011019283747, "p_correct_given_cut": 0.013422818791946308, "no_answer_finished_share": 0.033203125, "loop_end_share_of_cut": 0.8053691275167785, "tail_dup8_cut_mean": 0.751610067114094, "median_tokens_correct": 9659, "by_source": { "gsm8k": { "problems": 128, "samples": 256, "acc": 0.70703125, "acc_se": 0.0370076536586475, "cut_share": 0.125, "mean_tokens": 12543.30859375 }, "math": { "problems": 128, "samples": 256, "acc": 0.4765625, "acc_se": 0.040321731073965326, "cut_share": 0.45703125, "mean_tokens": 22915.7734375 } }, "run": { "name": "dev/a4", "model": "qwen35-a4", "data": "dev256", "n": 2, "shards": 1, "est": 8 }, "sampling": { "temperature": 0.6, "top_p": 0.95, "top_k": 20, "presence_penalty": 1.5, "repetition_penalty": 1.0, "max_tokens": 32768, "thinking": true }, "edit_note": { "name": "a4", "spec": { "about": "flip and triple the loop component", "layers": { "15": { "alpha": 4 }, "19": { "alpha": 4 } } }, "built": { "15": { "alpha": 4.0, "n_cols": 64, "dir": "released", "relative_change": 0.07090040296316147 }, "19": { "alpha": 4.0, "n_cols": 64, "dir": "released", "relative_change": 0.05403315648436546 } }, "edited_sha256": { "model.language_model.layers.15.mlp.down_proj.weight": "39afb7d3b64af312259abe3317192026bebd4c2f507e2312b38f97af902b41bc", "model.language_model.layers.19.mlp.down_proj.weight": "068a77ed4d5a57880064e29086367b150ea0fb5f2fb1ca2ea4b87caf14190467" } }, "tokens_total": 9077525, "gpu_seconds": 1359, "tok_per_s_per_gpu": 6911 }