{ "bench_version": 1, "track": "categorical_persistent_sequences", "model": "mlp_token_logits", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.003, "temp": 0.8 }, "sweep": [ { "cfg": { "lr": 0.001, "temp": 0.8 }, "mean": 0.49406249999999996 }, { "cfg": { "lr": 0.001, "temp": 1.0 }, "mean": 0.49406249999999996 }, { "cfg": { "lr": 0.001, "temp": 1.2 }, "mean": 0.49406249999999996 }, { "cfg": { "lr": 0.003, "temp": 0.8 }, "mean": 0.490625 }, { "cfg": { "lr": 0.003, "temp": 1.0 }, "mean": 0.490625 }, { "cfg": { "lr": 0.003, "temp": 1.2 }, "mean": 0.490625 }, { "cfg": { "lr": 0.01, "temp": 0.8 }, "mean": 0.49687499999999996 }, { "cfg": { "lr": 0.01, "temp": 1.0 }, "mean": 0.49687499999999996 }, { "cfg": { "lr": 0.01, "temp": 1.2 }, "mean": 0.49687499999999996 } ], "full": { "mean": 0.49445312500000005, "std": 0.015781830125475786, "per_seed": [ 0.460625, 0.498125, 0.514375, 0.489375, 0.483125, 0.50875, 0.501875, 0.499375 ], "n": 8 } }, "idea": { "mean": 0.48640625000000004, "std": 0.013064406260427612, "per_seed": [ 0.480625, 0.503125, 0.48625, 0.49875, 0.47375, 0.464375, 0.4825, 0.501875 ], "n": 8 }, "comparison": { "delta_mean": -0.008046875000000002, "idea_wins": 4, "n_pairs": 8, "per_seed_diffs": [ 0.020000000000000018, 0.00500000000000006, -0.02812500000000001, 0.009375000000000022, -0.009375000000000022, -0.04437500000000005, -0.019374999999999976, 0.0024999999999999467 ], "p_value": 0.3411, "mde": 0.01794619570947467, "mde_rel_pct": 3.629503951355282, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "custom_track": { "name": "categorical_persistent_sequences", "file": "categorical_track.py", "domain": "diffusion-sampling" }, "idea_config": { "lr": 0.01, "temp": 1.0, "rho": 0.0 }, "idea_sweep": [ { "cfg": { "lr": 0.001, "temp": 0.8, "rho": 0.0 }, "result": { "mean": 0.4915625, "std": 0.007335434121440939, "per_seed": [ 0.49375, 0.495, 0.488125, 0.475, 0.49, 0.495625, 0.501875, 0.493125 ], "n": 8 } }, { "cfg": { "lr": 0.001, "temp": 0.8, "rho": 0.5 }, "result": { "mean": 0.49539062500000003, "std": 0.012062408921288268, "per_seed": [ 0.49375, 0.523125, 0.493125, 0.49375, 0.4775, 0.49875, 0.495, 0.488125 ], "n": 8 } }, { "cfg": { "lr": 0.001, "temp": 0.8, "rho": 0.75 }, "result": { "mean": 0.495390625, "std": 0.011030574961640704, "per_seed": [ 0.49125, 0.515625, 0.493125, 0.498125, 0.47875, 0.5075, 0.485, 0.49375 ], "n": 8 } }, { "cfg": { "lr": 0.001, "temp": 1.0, "rho": 0.0 }, "result": { "mean": 0.49109375, "std": 0.007842131307080995, "per_seed": [ 0.493125, 0.495, 0.48625, 0.473125, 0.49125, 0.496875, 0.500625, 0.4925 ], "n": 8 } }, { "cfg": { "lr": 0.001, "temp": 1.0, "rho": 0.5 }, "result": { "mean": 0.495859375, "std": 0.012342513783033639, "per_seed": [ 0.4925, 0.52375, 0.49375, 0.495, 0.47625, 0.499375, 0.495625, 0.490625 ], "n": 8 } }, { "cfg": { "lr": 0.001, "temp": 1.0, "rho": 0.75 }, "result": { "mean": 0.49484375, "std": 0.011214135541248827, "per_seed": [ 0.491875, 0.515625, 0.493125, 0.49125, 0.47875, 0.50875, 0.485, 0.494375 ], "n": 8 } }, { "cfg": { "lr": 0.001, "temp": 1.2, "rho": 0.0 }, "result": { "mean": 0.49007812500000003, "std": 0.007351641371787327, "per_seed": [ 0.49125, 0.49625, 0.48625, 0.474375, 0.490625, 0.49375, 0.500625, 0.4875 ], "n": 8 } }, { "cfg": { "lr": 0.001, "temp": 1.2, "rho": 0.5 }, "result": { "mean": 0.49554687499999994, "std": 0.011727861041527343, "per_seed": [ 0.491875, 0.5225, 0.49375, 0.493125, 0.4775, 0.498125, 0.49625, 0.49125 ], "n": 8 } }, { "cfg": { "lr": 0.001, "temp": 1.2, "rho": 0.75 }, "result": { "mean": 0.49484375, "std": 0.01113548487662301, "per_seed": [ 0.49125, 0.515625, 0.495625, 0.491875, 0.478125, 0.5075, 0.485, 0.49375 ], "n": 8 } }, { "cfg": { "lr": 0.003, "temp": 0.8, "rho": 0.0 }, "result": { "mean": 0.48953125, "std": 0.005587985856952403, "per_seed": [ 0.491875, 0.4975, 0.49125, 0.486875, 0.4775, 0.490625, 0.486875, 0.49375 ], "n": 8 } }, { "cfg": { "lr": 0.003, "temp": 0.8, "rho": 0.5 }, "result": { "mean": 0.49203125000000003, "std": 0.013609904194648118, "per_seed": [ 0.503125, 0.51375, 0.48375, 0.496875, 0.464375, 0.495625, 0.490625, 0.488125 ], "n": 8 } }, { "cfg": { "lr": 0.003, "temp": 0.8, "rho": 0.75 }, "result": { "mean": 0.494921875, "std": 0.00963411449534284, "per_seed": [ 0.488125, 0.514375, 0.49, 0.49375, 0.483125, 0.505625, 0.48875, 0.495625 ], "n": 8 } }, { "cfg": { "lr": 0.003, "temp": 1.0, "rho": 0.0 }, "result": { "mean": 0.488984375, "std": 0.0071393598441579515, "per_seed": [ 0.494375, 0.496875, 0.49, 0.484375, 0.475625, 0.486875, 0.485, 0.49875 ], "n": 8 } }, { "cfg": { "lr": 0.003, "temp": 1.0, "rho": 0.5 }, "result": { "mean": 0.491328125, "std": 0.013193413588202828, "per_seed": [ 0.5025, 0.513125, 0.481875, 0.494375, 0.465, 0.491875, 0.49125, 0.490625 ], "n": 8 } }, { "cfg": { "lr": 0.003, "temp": 1.0, "rho": 0.75 }, "result": { "mean": 0.49421875, "std": 0.009456675469608747, "per_seed": [ 0.489375, 0.510625, 0.485, 0.495, 0.483125, 0.5075, 0.488125, 0.495 ], "n": 8 } }, { "cfg": { "lr": 0.003, "temp": 1.2, "rho": 0.0 }, "result": { "mean": 0.490234375, "std": 0.0072074282156935204, "per_seed": [ 0.495625, 0.5, 0.490625, 0.486875, 0.475, 0.490625, 0.486875, 0.49625 ], "n": 8 } }, { "cfg": { "lr": 0.003, "temp": 1.2, "rho": 0.5 }, "result": { "mean": 0.49203125000000003, "std": 0.013049447725382855, "per_seed": [ 0.50125, 0.5125, 0.48375, 0.495625, 0.464375, 0.493125, 0.495, 0.490625 ], "n": 8 } }, { "cfg": { "lr": 0.003, "temp": 1.2, "rho": 0.75 }, "result": { "mean": 0.494609375, "std": 0.010448616366982527, "per_seed": [ 0.494375, 0.51125, 0.4825, 0.495625, 0.48, 0.50875, 0.48875, 0.495625 ], "n": 8 } }, { "cfg": { "lr": 0.01, "temp": 0.8, "rho": 0.0 }, "result": { "mean": 0.487890625, "std": 0.01414882359012138, "per_seed": [ 0.48, 0.51, 0.488125, 0.503125, 0.4725, 0.468125, 0.48125, 0.5 ], "n": 8 } }, { "cfg": { "lr": 0.01, "temp": 0.8, "rho": 0.5 }, "result": { "mean": 0.494140625, "std": 0.014337374331947086, "per_seed": [ 0.49625, 0.525625, 0.485, 0.495, 0.47625, 0.48375, 0.4875, 0.50375 ], "n": 8 } }, { "cfg": { "lr": 0.01, "temp": 0.8, "rho": 0.75 }, "result": { "mean": 0.494375, "std": 0.012480453467322404, "per_seed": [ 0.48, 0.5225, 0.485, 0.49875, 0.490625, 0.490625, 0.48625, 0.50125 ], "n": 8 } }, { "cfg": { "lr": 0.01, "temp": 1.0, "rho": 0.0 }, "result": { "mean": 0.48640625000000004, "std": 0.013064406260427612, "per_seed": [ 0.480625, 0.503125, 0.48625, 0.49875, 0.47375, 0.464375, 0.4825, 0.501875 ], "n": 8 } }, { "cfg": { "lr": 0.01, "temp": 1.0, "rho": 0.5 }, "result": { "mean": 0.4946875, "std": 0.012751378601939463, "per_seed": [ 0.495625, 0.518125, 0.48625, 0.498125, 0.475625, 0.48625, 0.48875, 0.50875 ], "n": 8 } }, { "cfg": { "lr": 0.01, "temp": 1.0, "rho": 0.75 }, "result": { "mean": 0.495, "std": 0.010798220339481851, "per_seed": [ 0.481875, 0.518125, 0.48375, 0.4975, 0.49375, 0.493125, 0.489375, 0.5025 ], "n": 8 } }, { "cfg": { "lr": 0.01, "temp": 1.2, "rho": 0.0 }, "result": { "mean": 0.48726562500000004, "std": 0.012251185768503196, "per_seed": [ 0.483125, 0.499375, 0.489375, 0.495625, 0.47375, 0.464375, 0.489375, 0.503125 ], "n": 8 } }, { "cfg": { "lr": 0.01, "temp": 1.2, "rho": 0.5 }, "result": { "mean": 0.49382812499999995, "std": 0.013329645282578786, "per_seed": [ 0.49625, 0.516875, 0.485, 0.49375, 0.47375, 0.485625, 0.488125, 0.51125 ], "n": 8 } }, { "cfg": { "lr": 0.01, "temp": 1.2, "rho": 0.75 }, "result": { "mean": 0.49671875, "std": 0.010325512623473003, "per_seed": [ 0.4825, 0.51875, 0.486875, 0.495, 0.49625, 0.494375, 0.495625, 0.504375 ], "n": 8 } } ], "mechanism_signature": { "prediction": "Gaussian refresh sign correlation is 2 asin(rho)/pi", "rho": 0.75, "rows": [ { "predicted_sign_corr": 0.5398930876747683, "observed_sign_corr": 0.060386206896551727 }, { "predicted_sign_corr": 0.5398930876747683, "observed_sign_corr": 0.06482068965517242 }, { "predicted_sign_corr": 0.5398930876747683, "observed_sign_corr": 0.07049655172413793 }, { "predicted_sign_corr": 0.5398930876747683, "observed_sign_corr": 0.06498448275862069 } ], "mean_abs_error": 0.4747211049161476, "tolerance": 0.04, "confirmed": false } } }