{ "problems": 256, "samples": 512, "acc": 0.14453125, "acc_se": 0.018212198860840877, "cut_share": 0.873046875, "mean_tokens": 29486.103515625, "closed_share": 0.140625, "p_finish": 0.126953125, "p_correct_given_finished": 0.9846153846153847, "p_correct_given_cut": 0.02237136465324385, "no_answer_finished_share": 0.0, "loop_end_share_of_cut": 0.9731543624161074, "tail_dup8_cut_mean": 0.9167756152125279, "median_tokens_correct": 5525.5, "by_source": { "gsm8k": { "problems": 128, "samples": 256, "acc": 0.15234375, "acc_se": 0.025750660389962682, "cut_share": 0.875, "mean_tokens": 29301.203125 }, "math": { "problems": 128, "samples": 256, "acc": 0.13671875, "acc_se": 0.025843808317160873, "cut_share": 0.87109375, "mean_tokens": 29671.00390625 } }, "run": { "name": "dev/randdir", "model": "qwen35-randdir", "data": "dev256", "n": 2, "shards": 1, "est": 15 }, "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": "randdir", "spec": { "about": "control: same 64 neurons, same size of change, but a random direction instead of u", "layers": { "15": { "alpha": 6, "dir": "random_orth:15", "norm_match": true }, "19": { "alpha": 6, "dir": "random_orth:19", "norm_match": true } } }, "built": { "15": { "alpha": 26.746307373046875, "n_cols": 64, "dir": "random_orth:15", "relative_change": 0.10634875297546387 }, "19": { "alpha": 19.5016508102417, "n_cols": 64, "dir": "random_orth:19", "relative_change": 0.08105900138616562 } }, "edited_sha256": { "model.language_model.layers.15.mlp.down_proj.weight": "067f23efd0db9a0484345af96d5442378ae673e56cd3635fccaff65339edc3ae", "model.language_model.layers.19.mlp.down_proj.weight": "22070034b4c6e17f38b489194b95b7a1bbe1fb820d5495b2b3bfd6fb05b0928e" } }, "tokens_total": 15096885, "gpu_seconds": 2681, "tok_per_s_per_gpu": 5752 }