{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01, "weight_decay": 0.0 }, "sweep": [ { "cfg": { "lr": 0.001, "weight_decay": 0.0 }, "mean": 0.0009722943214001134 }, { "cfg": { "lr": 0.001, "weight_decay": 0.0001 }, "mean": 0.0009676419722381979 }, { "cfg": { "lr": 0.001, "weight_decay": 0.001 }, "mean": 0.0009444169700145721 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0 }, "mean": 0.0004595852587954141 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001 }, "mean": 0.000500455847941339 }, { "cfg": { "lr": 0.003, "weight_decay": 0.001 }, "mean": 0.0005452305340440944 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0 }, "mean": 0.00011186753999936627 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0001 }, "mean": 0.0001743265111144865 }, { "cfg": { "lr": 0.01, "weight_decay": 0.001 }, "mean": 0.0002742330798355397 } ], "full": { "mean": 8.200101638067281e-05, "std": 6.057323033203833e-05, "per_seed": [ 0.0002028812450589612, 0.00016224317369051278, 4.15411705034785e-05, 4.08045707445126e-05, 5.192989920033142e-05, 2.7118352591060102e-05, 4.964483741787262e-05, 7.984488183865324e-05 ], "n": 8 }, "idea_grid": [ { "cfg": { "lr": 0.001, "c": 2.0 }, "mean": 0.0009725104464450851 }, { "cfg": { "lr": 0.001, "c": 5.0 }, "mean": 0.0009728349541546777 }, { "cfg": { "lr": 0.001, "c": 10.0 }, "mean": 0.0009733773185871542 }, { "cfg": { "lr": 0.003, "c": 2.0 }, "mean": 0.0004606272341334261 }, { "cfg": { "lr": 0.003, "c": 5.0 }, "mean": 0.0004622471169568598 }, { "cfg": { "lr": 0.003, "c": 10.0 }, "mean": 0.00046506051876349375 }, { "cfg": { "lr": 0.01, "c": 2.0 }, "mean": 0.0001056715045706369 }, { "cfg": { "lr": 0.01, "c": 5.0 }, "mean": 9.620005857868819e-05 }, { "cfg": { "lr": 0.01, "c": 10.0 }, "mean": 8.880743644112954e-05 } ], "idea_best_cfg": { "lr": 0.01, "c": 10.0 } }, "idea": { "mean": 6.680865521957458e-05, "std": 4.834767024791531e-05, "per_seed": [ 0.00018218360492028296, 9.848197805695236e-05, 3.606851532822475e-05, 3.849564745905809e-05, 5.274467548588291e-05, 2.1851565179531462e-05, 5.2063427574466914e-05, 5.257982775219716e-05 ], "n": 8 }, "comparison": { "delta_mean": -1.5192361161098233e-05, "idea_wins": 6, "n_pairs": 8, "per_seed_diffs": [ -2.0697640138678253e-05, -6.376119563356042e-05, -5.472655175253749e-06, -2.3089232854545116e-06, 8.147762855514884e-07, -5.2667874115286395e-06, 2.4185901565942913e-06, -2.7265054086456075e-05 ], "p_value": 0.04715, "mde": 1.8590723573088402e-05, "mde_rel_pct": 22.671333104927385, "verdict": "idea better (significant)", "system_worked": true }, "mechanism_signature": { "mechanism_signature": { "E_star": 0.01, "tau": 0.005, "c": 10.0, "transition_width_10_to_90": 0.0219722457735, "observed": [ { "seed": 0, "max_energy": 0.6635691743140342, "mean_gate": 0.2657892433072299, "active_fraction": 0.15333333333333332, "mean_abs_grad_work": 0.009355158761435146, "mean_damping_energy_term": 0.3849997261763783, "bound_proxy_B_over_2c": 0.0015436692659204674 }, { "seed": 1, "max_energy": 0.6644394179311348, "mean_gate": 0.27762200195231984, "active_fraction": 0.16, "mean_abs_grad_work": 0.010108855998219184, "mean_damping_energy_term": 0.4196295187630554, "bound_proxy_B_over_2c": 0.0020301451462910263 }, { "seed": 2, "max_energy": 0.6638438488153042, "mean_gate": 0.2998147450677648, "active_fraction": 0.21333333333333335, "mean_abs_grad_work": 0.009347904387412672, "mean_damping_energy_term": 0.4398196667126576, "bound_proxy_B_over_2c": 0.001223365468613331 }, { "seed": 3, "max_energy": 0.6624586769175949, "mean_gate": 0.27958306215077783, "active_fraction": 0.16, "mean_abs_grad_work": 0.009035967432046198, "mean_damping_energy_term": 0.4271317061643127, "bound_proxy_B_over_2c": 0.0013104913539427757 }, { "seed": 4, "max_energy": 0.6631062422602554, "mean_gate": 0.27506431785991164, "active_fraction": 0.17333333333333334, "mean_abs_grad_work": 0.008536183482930577, "mean_damping_energy_term": 0.40087036121829356, "bound_proxy_B_over_2c": 0.00110232781849845 }, { "seed": 5, "max_energy": 0.6644941904560255, "mean_gate": 0.2890626448038116, "active_fraction": 0.19333333333333333, "mean_abs_grad_work": 0.010214908040596604, "mean_damping_energy_term": 0.4445307720323369, "bound_proxy_B_over_2c": 0.001505933901644309 }, { "seed": 6, "max_energy": 0.6638318966870429, "mean_gate": 0.2855342704509694, "active_fraction": 0.17333333333333334, "mean_abs_grad_work": 0.009050130401310334, "mean_damping_energy_term": 0.4042768311227779, "bound_proxy_B_over_2c": 0.0010688055686387085 }, { "seed": 7, "max_energy": 0.664137040348578, "mean_gate": 0.26710557821443215, "active_fraction": 0.15333333333333332, "mean_abs_grad_work": 0.00960955584816272, "mean_damping_energy_term": 0.4161792169876806, "bound_proxy_B_over_2c": 0.001771787163452233 } ], "prediction": "higher energy produces higher q and nonzero dissipative term 2*c*q*E", "confirmed": true } }, "custom_track": null }