{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01 }, "sweep": [ { "cfg": { "lr": 0.001 }, "mean": 0.005440622277092189 }, { "cfg": { "lr": 0.003 }, "mean": 0.0012262430100236088 }, { "cfg": { "lr": 0.01 }, "mean": 0.000702349963830784 } ], "full": { "mean": 0.0006528829871967901, "std": 0.0001692890895210604, "per_seed": [ 0.0006883637397550046, 0.00054119125707075, 0.0005475711077451706, 0.0010322737507522106, 0.00068665825529024, 0.00041018708725459874, 0.0006441730074584484, 0.0006726456922478974 ], "n": 8 } }, "idea": { "mean": 0.000652743416139856, "std": 0.0001692304843358837, "per_seed": [ 0.0006887014023959637, 0.0005411907332018018, 0.0005468056187964976, 0.0010318164713680744, 0.0006858076085336506, 0.00041013583540916443, 0.0006440518773160875, 0.0006734377820976079 ], "n": 8 }, "comparison": { "delta_mean": -1.3957105693407357e-07, "idea_wins": 6, "n_pairs": 8, "per_seed_diffs": [ 3.3766264095902443e-07, -5.238689482212067e-10, -7.654889486730099e-07, -4.5727938413619995e-07, -8.506467565894127e-07, -5.125184543430805e-08, -1.2113014236092567e-07, 7.920898497104645e-07 ], "p_value": 0.45895, "mde": 4.6032512964596226e-07, "mde_rel_pct": 0.07050652853161457, "verdict": "no measurable effect", "system_worked": false }, "mechanism_signature": { "mechanism_signature": { "carrier_amp": 0.3, "predicted_abs_delta_mean": 0.0008173562528099865, "observed_nominal_abs_delta_mean": 5.861693352926522e-10, "predicted_observed_correlation_mean": NaN, "quadratic_prediction": "not confirmed at NN scale: only one amplitude was behaviorally tested", "confirmed": false }, "idea_sweep": [ { "amp": 0.05, "result": { "mean": 0.0006528785997943487, "std": 0.0001692863867624003, "per_seed": [ 0.0006883704336360097, 0.0005411913734860718, 0.0005475485231727362, 0.001032259315252304, 0.000686634739395231, 0.00041018894989974797, 0.0006441685254685581, 0.0006726669380441308 ], "n": 8 } }, { "amp": 0.15, "result": { "mean": 0.0006528485828312114, "std": 0.00016926723045994872, "per_seed": [ 0.000688436848577112, 0.0005411922466009855, 0.0005473769851960242, 0.0010321518639102578, 0.0006864455062896013, 0.0004102009115740657, 0.0006441408768296242, 0.0006728434236720204 ], "n": 8 } }, { "amp": 0.3, "result": { "mean": 0.000652743416139856, "std": 0.0001692304843358837, "per_seed": [ 0.0006887014023959637, 0.0005411907332018018, 0.0005468056187964976, 0.0010318164713680744, 0.0006858076085336506, 0.00041013583540916443, 0.0006440518773160875, 0.0006734377820976079 ], "n": 8 } } ], "track_justification": "Dynamics matches the proposed hidden-state observability/reachability and stability/control structure." }, "idea_sweep": [ { "amp": 0.05, "result": { "mean": 0.0006528785997943487, "std": 0.0001692863867624003, "per_seed": [ 0.0006883704336360097, 0.0005411913734860718, 0.0005475485231727362, 0.001032259315252304, 0.000686634739395231, 0.00041018894989974797, 0.0006441685254685581, 0.0006726669380441308 ], "n": 8 } }, { "amp": 0.15, "result": { "mean": 0.0006528485828312114, "std": 0.00016926723045994872, "per_seed": [ 0.000688436848577112, 0.0005411922466009855, 0.0005473769851960242, 0.0010321518639102578, 0.0006864455062896013, 0.0004102009115740657, 0.0006441408768296242, 0.0006728434236720204 ], "n": 8 } }, { "amp": 0.3, "result": { "mean": 0.000652743416139856, "std": 0.0001692304843358837, "per_seed": [ 0.0006887014023959637, 0.0005411907332018018, 0.0005468056187964976, 0.0010318164713680744, 0.0006858076085336506, 0.00041013583540916443, 0.0006440518773160875, 0.0006734377820976079 ], "n": 8 } } ] }