| { |
| "problems": 256, |
| "samples": 512, |
| "acc": 0.044921875, |
| "acc_se": 0.009371648451158184, |
| "cut_share": 0.990234375, |
| "mean_tokens": 32472.140625, |
| "closed_share": 0.009765625, |
| "p_finish": 0.009765625, |
| "p_correct_given_finished": 1.0, |
| "p_correct_given_cut": 0.03550295857988166, |
| "no_answer_finished_share": 0.0, |
| "loop_end_share_of_cut": 0.9842209072978304, |
| "tail_dup8_cut_mean": 0.9093725838264299, |
| "median_tokens_correct": 32768, |
| "by_source": { |
| "gsm8k": { |
| "problems": 128, |
| "samples": 256, |
| "acc": 0.0625, |
| "acc_se": 0.015686395128453965, |
| "cut_share": 0.984375, |
| "mean_tokens": 32279.65234375 |
| }, |
| "math": { |
| "problems": 128, |
| "samples": 256, |
| "acc": 0.02734375, |
| "acc_se": 0.010087879347602812, |
| "cut_share": 0.99609375, |
| "mean_tokens": 32664.62890625 |
| } |
| }, |
| "run": { |
| "name": "dev/base+pp0", |
| "model": "base", |
| "data": "dev256", |
| "n": 2, |
| "shards": 1, |
| "est": 15, |
| "set": [ |
| "presence_penalty=0" |
| ] |
| }, |
| "sampling": { |
| "temperature": 0.6, |
| "top_p": 0.95, |
| "top_k": 20, |
| "presence_penalty": 0.0, |
| "repetition_penalty": 1.0, |
| "max_tokens": 32768, |
| "thinking": true |
| }, |
| "edit_note": null, |
| "tokens_total": 16625736, |
| "gpu_seconds": 2934, |
| "tok_per_s_per_gpu": 5753 |
| } |
