{ "device": "cuda", "seed": 275, "math_check": { "predictor_max_abs_error": 5.960464477539063e-08, "noise_variance": 0.017955534160137177, "noise_target": 0.018199999999999997, "relative_variance_error": 0.013432189003451701 }, "baseline": { "test_mse": 0.00013550324365496635, "train_mse": 6.543769268319011e-05, "checkpoints": [ 0.8140243291854858, 0.006550719495862722, 0.0005533923977054656, 0.00013550324365496635 ], "alignment_mean": 0.8967403173446655, "aligned_fraction": 0.875, "mode_counts": [ 42, 22 ], "weight_diversity": 1.1638377904891968, "seconds": 1.338935375213623 }, "particle": { "test_mse": 0.1158861592411995, "train_mse": 0.09725860506296158, "checkpoints": [ 0.8507484793663025, 0.5212430357933044, 0.7502710223197937, 0.1158861592411995 ], "alignment_mean": 0.2828409671783447, "aligned_fraction": 0.0, "mode_counts": [ 28, 36 ], "weight_diversity": 1.40092933177948, "seconds": 1.3214137554168701 }, "config": { "width": 64, "dimension": 20, "steps": 700, "batch": 96, "schedule": "lambda 1.0 -> 0.01 cosine", "teacher": "two-index ReLU regression" } }