{ "bench_version": 1, "track": "robust_cbf_pendulum_policy", "model": "local_mlp_tiny", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.006, "eps": 0.0, "kc": 1.5, "epochs": 15 }, "sweep": [ { "cfg": { "lr": 0.001, "eps": 0.0, "kc": 1.5, "epochs": 15 }, "mean": 0.021794200874865055 }, { "cfg": { "lr": 0.003, "eps": 0.0, "kc": 1.5, "epochs": 15 }, "mean": 0.002334954886464402 }, { "cfg": { "lr": 0.006, "eps": 0.0, "kc": 1.5, "epochs": 15 }, "mean": 0.0018331437022425234 } ], "full": { "mean": 0.0019978795608039945, "std": 0.0005495084055684182, "per_seed": [ 0.0015628907131031156, 0.0015461408765986562, 0.0021467802580446005, 0.0020767629612237215, 0.0030877823010087013, 0.0011752876453101635, 0.0020312685519456863, 0.0023561231791973114 ], "n": 8 } }, "idea": { "mean": 0.667444035410881, "std": 0.05083093845847301, "per_seed": [ 0.7455176115036011, 0.6221279501914978, 0.5776458382606506, 0.6847237944602966, 0.7122350931167603, 0.6813809871673584, 0.6877815127372742, 0.6281394958496094 ], "n": 8, "selection_sweep": [ { "cfg": { "lr": 0.001, "eps": 0.05, "kc": 1.5, "epochs": 15 }, "mean": 0.6575037986040115 }, { "cfg": { "lr": 0.003, "eps": 0.15, "kc": 1.5, "epochs": 15 }, "mean": 0.6907831430435181 }, { "cfg": { "lr": 0.006, "eps": 0.3, "kc": 1.5, "epochs": 15 }, "mean": 0.7713967263698578 } ] }, "comparison": { "delta_mean": 0.665446155850077, "idea_wins": 0, "n_pairs": 8, "per_seed_diffs": [ 0.743954720790498, 0.6205818093148991, 0.575499058002606, 0.6826470314990729, 0.7091473108157516, 0.6802056995220482, 0.6857502441853285, 0.6257833726704121 ], "p_value": 0.0081, "mde": 0.04544311646025259, "mde_rel_pct": 2274.567363908823, "verdict": "idea worse (significant)", "system_worked": false }, "mechanism_signature": { "prediction": { "uncertainty_boundary_shift_per_epsilon": 1.0, "virtual_shift_formula": "Delta/kc" }, "observed_from_trained_models": { "shift_per_epsilon_mean": 0.9999994114041328, "mean_intervention": 0.3867955468595028, "feasible_fraction_mean": 0.9265625, "n_models": 8 }, "confirmed": true }, "custom_track": { "name": "robust_cbf_pendulum_policy", "file": "robust_cbf_pendulum_policy.py", "domain": "dynamics" } }