{ "math_verification": { "log_ratio_sweep": [ { "r": 0.1, "predicted_log_ratio": 0.1, "observed_log_ratio": 0.09997299621889233, "relative_error": 0.00027003781107676206 }, { "r": 0.25, "predicted_log_ratio": 0.25, "observed_log_ratio": 0.2498593046380371, "relative_error": 0.0005627814478516013 }, { "r": 0.5, "predicted_log_ratio": 0.5, "observed_log_ratio": 0.49962443651370425, "relative_error": 0.0007511269725914982 }, { "r": 0.75, "predicted_log_ratio": 0.75, "observed_log_ratio": 0.749576855177982, "relative_error": 0.0005641930960239941 }, { "r": 1.0, "predicted_log_ratio": 1.0, "observed_log_ratio": 1.0, "relative_error": 0.0 } ], "schedule_cumulative": { "predicted_log_shrinkage": -0.6727382872123923, "observed_log_shrinkage": -0.6727382872123935, "absolute_error": 1.2212453270876722e-15, "constant_log_shrinkage": -0.9050205413161395 }, "lambda_zero_sweep": [ { "r": 0.1, "absolute_difference": 0.0 }, { "r": 0.5, "absolute_difference": 0.0 }, { "r": 1.0, "absolute_difference": 0.0 } ], "lambda_scaling_sweep": [ { "lambda": 0.0, "observed_log_shrinkage": 0.0, "predicted_first_order": -0.0, "normalized_observed": 0.0, "predicted_slope": -3.3564166666666666 }, { "lambda": 0.02, "observed_log_shrinkage": -0.06714283992963532, "predicted_first_order": -0.06712833333333333, "normalized_observed": -3.357141996481766, "predicted_slope": -3.3564166666666666 }, { "lambda": 0.05, "observed_log_shrinkage": -0.1679115441609461, "predicted_first_order": -0.16782083333333334, "normalized_observed": -3.3582308832189223, "predicted_slope": -3.3564166666666666 }, { "lambda": 0.1, "observed_log_shrinkage": -0.33600480778021485, "predicted_first_order": -0.33564166666666667, "normalized_observed": -3.360048077802148, "predicted_slope": -3.3564166666666666 }, { "lambda": 0.2, "observed_log_shrinkage": -0.6727382872123935, "predicted_first_order": -0.6712833333333333, "normalized_observed": -3.363691436061967, "predicted_slope": -3.3564166666666666 }, { "lambda": 0.4, "observed_log_shrinkage": -1.348405710433359, "predicted_first_order": -1.3425666666666667, "normalized_observed": -3.3710142760833977, "predicted_slope": -3.3564166666666666 } ] }, "mini_experiment": { "baseline": { "final_loss": { "mean": 0.058479504038890205, "std": 0.024347165163437805 }, "loss_at_100": { "mean": 0.7814053297042847, "std": 0.3252069081099237 }, "loss_at_300": { "mean": 0.07287957891821861, "std": 0.03459615379637844 }, "final_norm": { "mean": 3.473102251688639, "std": 0.18855447223524782 }, "norm_at_300": { "mean": 3.4431944688161216, "std": 0.19383601011930413 } }, "idea": { "final_loss": { "mean": 0.028985902046163876, "std": 0.009098264729836133 }, "loss_at_100": { "mean": 0.7748445272445679, "std": 0.3224545840341833 }, "loss_at_300": { "mean": 0.04409672816594442, "std": 0.020935670726080508 }, "final_norm": { "mean": 3.5797719955444336, "std": 0.1882472925523988 }, "norm_at_300": { "mean": 3.521018107732137, "std": 0.1946352067112015 } }, "seeds": [ 1333, 1334, 1335 ] }, "settings": { "steps": 500, "peak_lr": 0.025, "nominal_decay": 0.08, "data": "fixed synthetic linear regression", "seed": 1333 } }