{ "problems": 256, "samples": 512, "acc": 0.517578125, "acc_se": 0.02840503248849555, "cut_share": 0.123046875, "mean_tokens": 13514.556640625, "closed_share": 0.84375, "p_finish": 0.876953125, "p_correct_given_finished": 0.5902004454342984, "p_correct_given_cut": 0.0, "no_answer_finished_share": 0.09375, "loop_end_share_of_cut": 0.09523809523809523, "tail_dup8_cut_mean": 0.13634126984126982, "median_tokens_correct": 5799, "by_source": { "gsm8k": { "problems": 128, "samples": 256, "acc": 0.5546875, "acc_se": 0.03891458214466775, "cut_share": 0.0078125, "mean_tokens": 8373.234375 }, "math": { "problems": 128, "samples": 256, "acc": 0.48046875, "acc_se": 0.041280241148031895, "cut_share": 0.23828125, "mean_tokens": 18655.87890625 } }, "run": { "name": "dev/a12", "model": "qwen35-a12", "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": "a12", "spec": { "about": "much stronger than released: flip and multiply by 11", "layers": { "15": { "alpha": 12 }, "19": { "alpha": 12 } } }, "built": { "15": { "alpha": 12.0, "n_cols": 64, "dir": "released", "relative_change": 0.21269792318344116 }, "19": { "alpha": 12.0, "n_cols": 64, "dir": "released", "relative_change": 0.16210497915744781 } }, "edited_sha256": { "model.language_model.layers.15.mlp.down_proj.weight": "175cfb4666e2a669bc9aed7ca8d6cd21868b50c3dcc61ee774ef24ec1886a842", "model.language_model.layers.19.mlp.down_proj.weight": "2f7a5abca079d8c61ec5e7fc0be48e0f899e707e5372f538b5759dade0624464" } }, "tokens_total": 6919453, "gpu_seconds": 1010, "tok_per_s_per_gpu": 7291 }