{ "bench_version": 1, "track": "custom:bidirectional_conditional_compatibility", "model": "two_conditional_mlp", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "sweep": { "best_cfg": { "lr": 0.01 }, "sweep": [ { "cfg": { "lr": 0.01 }, "mean": 1.326949805021286 }, { "cfg": { "lr": 0.03 }, "mean": 1.3306202590465546 }, { "cfg": { "lr": 0.1 }, "mean": 1.3483841717243195 } ], "full": { "mean": 1.3114151656627655, "std": 0.02693886115100743, "per_seed": [ 1.329000473022461, 1.3154711723327637, 1.3227543830871582, 1.3405731916427612, 1.3417959213256836, 1.3014893531799316, 1.2759835720062256, 1.2642532587051392 ], "n": 8 } }, "best_config": { "lr": 0.01 }, "full": { "per_seed": [ 1.329000473022461, 1.3154711723327637, 1.3227543830871582, 1.3405731916427612, 1.3417959213256836, 1.3014893531799316, 1.2759835720062256, 1.2642532587051392 ], "mean": 1.3114151656627655, "std": 0.028798853947698433 } }, "idea": { "per_seed": [ 1.328962802886963, 1.3155274391174316, 1.3226888179779053, 1.3404121398925781, 1.341781735420227, 1.3010740280151367, 1.2759394645690918, 1.2640597820281982 ], "mean": 1.3113057762384415, "std": 0.028841388419379284 }, "comparison": { "delta_mean": -0.00010938942432403564, "idea_wins": 7, "n_pairs": 8, "per_seed_diffs": [ -3.7670135498046875e-05, 5.626678466796875e-05, -6.556510925292969e-05, -0.00016105175018310547, -1.4185905456542969e-05, -0.0004153251647949219, -4.410743713378906e-05, -0.00019347667694091797 ], "p_value": 0.04555, "mde": 0.00012287599741212592, "mde_rel_pct": 0.009369725212078557, "verdict": "idea better (significant)", "system_worked": true }, "mechanism_signature": { "custom_track": { "name": "bidirectional_conditional_compatibility", "file": "cycle_bench.py", "domain": "paired categorical conditionals" }, "prediction": "cycle penalty lowers held-out 95th-percentile |Delta| while retaining task NLL", "baseline_heldout_p95": 0.06756381243467333, "idea_heldout_p95": 0.036757445335388196, "baseline_test_nll": 1.3114151656627655, "idea_test_nll": 1.3113057762384415, "observed_empirical_cycle_mean": 2.187573233669582e-07, "confirmed": true } }