{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.006, "epochs": 18, "wd": 0.0001 }, "sweep": [ { "cfg": { "lr": 0.0015, "epochs": 18, "wd": 0.0 }, "mean": 0.0016401863540522754 }, { "cfg": { "lr": 0.0015, "epochs": 18, "wd": 0.0001 }, "mean": 0.0016229975590249524 }, { "cfg": { "lr": 0.003, "epochs": 18, "wd": 0.0 }, "mean": 0.00105226032610517 }, { "cfg": { "lr": 0.003, "epochs": 18, "wd": 0.0001 }, "mean": 0.0010288678167853504 }, { "cfg": { "lr": 0.006, "epochs": 18, "wd": 0.0 }, "mean": 0.0007705322786932811 }, { "cfg": { "lr": 0.006, "epochs": 18, "wd": 0.0001 }, "mean": 0.0007483311783289537 } ], "full": { "mean": 0.0007712390506640077, "std": 0.00015769195775709773, "per_seed": [ 0.0006971143302507699, 0.0008527055033482611, 0.0005140206776559353, 0.0009294842020608485, 0.0005480053368955851, 0.0008613452082499862, 0.0008049127063713968, 0.0009623244404792786 ], "n": 8 } }, "idea": { "mean": 0.0007767367897031363, "std": 0.00015170973680853395, "per_seed": [ 0.0007888000109232962, 0.0009194124140776694, 0.0004823579511139542, 0.0008317303727380931, 0.0005752710276283324, 0.0008502130513079464, 0.0008313465514220297, 0.000934762938413769 ], "n": 8 }, "comparison": { "delta_mean": 5.497739039128646e-06, "idea_wins": 4, "n_pairs": 8, "per_seed_diffs": [ 9.168568067252636e-05, 6.670691072940826e-05, -3.16627265419811e-05, -9.775382932275534e-05, 2.7265690732747316e-05, -1.1132156942039728e-05, 2.6433845050632954e-05, -2.7561502065509558e-05 ], "p_value": 0.8404, "mde": 5.046532793104354e-05, "mde_rel_pct": 6.5434093213504685, "verdict": "no measurable effect", "system_worked": false }, "mechanism_signature": { "predicted_vs_observed": { "final_tau": 0.7026815647163405, "root_trace": [ [ 0.41413663615422736, -3.976707951927841 ], [ 0.5210281416079688, -1.3861438181717687 ], [ 0.5849097575396691, -0.8485201991462532 ], [ 0.6324862316350907, -0.6447059261927688 ], [ 0.6707034118207508, -0.5277147523207517 ], [ 0.7026815647163405, -0.44972691119487185 ] ], "observed_sensitivity_peak_token": 6, "observed_sensitivity_profile": [ 0.04500379413366318, 0.06951507180929184, 0.10417859256267548, 0.14784090220928192, 0.20971933007240295, 0.2682212293148041, 0.31737804412841797, 0.09851673245429993 ], "predicted_tau_c": 0.6045997880780726, "predicted_omega": 1.7320508075688772, "observed_final_root": -0.39328237958536605, "confirmed": true, "trained_model_probe_seed": 0 }, "note": "root schedule and test-time perturbation-compatible delayed system measured on trained benchmark model" }, "idea_sweep": [ { "cfg": { "lr": 0.006, "wd": 0.0001, "epochs": 18, "tau0": 0.1, "gamma": 0.08, "target": -0.05 }, "mean": 0.0007767367897031363, "per_seed": [ 0.0007888000109232962, 0.0009194124140776694, 0.0004823579511139542, 0.0008317303727380931, 0.0005752710276283324, 0.0008502130513079464, 0.0008313465514220297, 0.000934762938413769 ] }, { "cfg": { "lr": 0.006, "wd": 0.0001, "epochs": 18, "tau0": 0.6, "gamma": 0.08, "target": -0.05 }, "mean": 0.0007824423082638532, "per_seed": [ 0.0008003599941730499, 0.0009204379166476429, 0.0004860293702222407, 0.000844792346470058, 0.0005833087489008904, 0.0008501522243022919, 0.0007866035448387265, 0.0009878543205559254 ] }, { "cfg": { "lr": 0.006, "wd": 0.0001, "epochs": 18, "tau0": 0.6, "gamma": 0.16, "target": -0.05 }, "mean": 0.0007843925777706318, "per_seed": [ 0.0008042265544645488, 0.0009266391862183809, 0.0004910334246233106, 0.0008239991730079055, 0.0005916950758546591, 0.0008566270698793232, 0.0007817440782673657, 0.0009991760598495603 ] } ], "custom_track": null }