{ "seed": 1430, "verification": { "formula_sweeps": [ { "prediction": "uniform soft count", "m": 4, "observed": 3.310546875, "predicted": 3.310546875, "abs_error": 0.0 }, { "prediction": "uniform soft count", "m": 8, "observed": 5.251128673553467, "predicted": 5.251128673553467, "abs_error": 0.0 }, { "prediction": "uniform soft count", "m": 16, "observed": 7.055462837219238, "predicted": 7.05546330383001, "abs_error": 4.6661077135468076e-07 }, { "prediction": "uniform soft count", "m": 32, "observed": 7.888481140136719, "predicted": 7.888481303698535, "abs_error": 1.6356181653520707e-07 }, { "prediction": "q-expert finite-group soft count", "q": 1, "observed": 1.0, "predicted": 1.0, "abs_error": 0.0 }, { "prediction": "q-expert finite-group soft count", "q": 2, "observed": 1.999969482421875, "predicted": 1.999969482421875, "abs_error": 0.0 }, { "prediction": "q-expert finite-group soft count", "q": 4, "observed": 3.959909677505493, "predicted": 3.959909616969526, "abs_error": 6.05359673500061e-08 }, { "prediction": "q-expert finite-group soft count", "q": 8, "observed": 7.055462837219238, "predicted": 7.05546330383001, "abs_error": 4.6661077135468076e-07 } ], "lambda_sweep": [ { "lambda": 0, "observed": 0.0, "predicted": 0.0 }, { "lambda": 0.1, "observed": 0.18027440309524537, "predicted": 0.18027440309524537 }, { "lambda": 0.5, "observed": 0.9013720154762268, "predicted": 0.9013720154762268 }, { "lambda": 1, "observed": 1.8027440309524536, "predicted": 1.8027440309524536 }, { "lambda": 2, "observed": 3.6054880619049072, "predicted": 3.6054880619049072 }, { "lambda": 5, "observed": 9.013720154762268, "predicted": 9.013720154762268 } ], "epsilon_threshold_sweep": [ { "epsilon": 1.0536396232851564, "predicted_hinge": 0.0, "observed_hinge": 0.0 }, { "epsilon": 2.1072792465703127, "predicted_hinge": 0.0, "observed_hinge": 0.0 }, { "epsilon": 3.160918869855469, "predicted_hinge": 6.893588086553852, "observed_hinge": 6.893588086553852 } ], "epsilon_star": 2.1072792465703127, "base_penalty": 1.8027440309524536 }, "baseline": { "cross_entropy": 0.0012828032486140728, "max_hard_load": 32, "min_hard_load": 0, "load_std": 16.0, "dropped_fraction": 0.5, "mean_soft_expansion_penalty": 4.3349175453186035, "violation_fraction": 1.0, "mean_hard_neighborhood": 3.8958333333333335, "loss_trace": [ 2.077996015548706, 0.1533023864030838, 0.15128406882286072 ] }, "expansion_balanced": { "cross_entropy": 0.18213726580142975, "max_hard_load": 32, "min_hard_load": 0, "load_std": 13.076696830622021, "dropped_fraction": 0.390625, "mean_soft_expansion_penalty": 0.7138320803642273, "violation_fraction": 0.2916666666666667, "mean_hard_neighborhood": 6.0, "loss_trace": [ 2.077996015548706, 0.8379582166671753, 0.8262269496917725 ] }, "notes": "n=128, E=8, top-1 diagnostic, capacity=16/expert, identical Adam setup; expansion uses fixed random subset schedule." }