{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01, "weight_decay": 0.0001 }, "sweep": [ { "cfg": { "lr": 0.001, "weight_decay": 0.0 }, "mean": 0.001921573159052059 }, { "cfg": { "lr": 0.001, "weight_decay": 0.0001 }, "mean": 0.0018964317569043487 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0 }, "mean": 0.0011179699067724869 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001 }, "mean": 0.0010905505041591823 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0 }, "mean": 0.0006299250526353717 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0001 }, "mean": 0.0005868283187737688 } ], "full": { "mean": 0.0005577069205173757, "std": 0.00016235484867295895, "per_seed": [ 0.00046681062667630613, 0.0004913305747322738, 0.0004835708241444081, 0.0009056012495420873, 0.0005242845509201288, 0.0003316242655273527, 0.0006971154361963272, 0.0005613178364001215 ], "n": 8 } }, "idea": { "mean": 0.0005104744304844644, "std": 0.00016259431597973005, "per_seed": [ 0.0005603914032690227, 0.0004172833578195423, 0.00043428907520137727, 0.0008996041724458337, 0.00048453747876919806, 0.00045123344170860946, 0.000313605327391997, 0.0005228511872701347 ], "n": 8, "chosen_cfg": { "lr": 0.01, "weight_decay": 0.0001, "delta": 0.5, "beta": 0.7, "slope": 8.0 }, "sweep": [ { "cfg": { "lr": 0.001, "weight_decay": 0.0001, "delta": 0.25, "beta": 0.7, "slope": 8.0 }, "mean": 0.001631094521144405 }, { "cfg": { "lr": 0.001, "weight_decay": 0.0001, "delta": 0.35, "beta": 0.7, "slope": 8.0 }, "mean": 0.0015781476977281272 }, { "cfg": { "lr": 0.001, "weight_decay": 0.0001, "delta": 0.5, "beta": 0.7, "slope": 8.0 }, "mean": 0.0017034276097547263 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001, "delta": 0.25, "beta": 0.7, "slope": 8.0 }, "mean": 0.0012407805188558996 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001, "delta": 0.35, "beta": 0.7, "slope": 8.0 }, "mean": 0.0012196671304991469 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001, "delta": 0.5, "beta": 0.7, "slope": 8.0 }, "mean": 0.0011900454992428422 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0001, "delta": 0.25, "beta": 0.7, "slope": 8.0 }, "mean": 0.0006524444761453196 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0001, "delta": 0.35, "beta": 0.7, "slope": 8.0 }, "mean": 0.0006070969975553453 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0001, "delta": 0.5, "beta": 0.7, "slope": 8.0 }, "mean": 0.000577892002183944 } ] }, "comparison": { "delta_mean": -4.72324900329113e-05, "idea_wins": 6, "n_pairs": 8, "per_seed_diffs": [ 9.358077659271657e-05, -7.404721691273153e-05, -4.9281748943030834e-05, -5.9970770962536335e-06, -3.974707215093076e-05, 0.00011960917618125677, -0.00038351010880433023, -3.846664912998676e-05 ], "p_value": 0.5225, "mde": 0.00012771297744744513, "mde_rel_pct": 22.89965800118956, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "prediction": "expansive temporal receptive fields should yield r>1+delta and gate<0.5", "observed_mean_ratio": 1.284388935357143, "observed_mean_gate": 0.8395185023546219, "predicted_threshold": 1.35, "confirmed": false }, "protocol_note": "Baseline sweep and idea sweep use shared lr union; 8 paired seeds; dynamics is structurally matched to stability/control." }