{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.006, "weight_decay": 0.0001 }, "sweep": [ { "cfg": { "lr": 0.0015, "weight_decay": 0.0 }, "mean": 0.0016401863540522754 }, { "cfg": { "lr": 0.0015, "weight_decay": 0.0001 }, "mean": 0.0016229975590249524 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0 }, "mean": 0.00105226032610517 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001 }, "mean": 0.0010288678167853504 }, { "cfg": { "lr": 0.006, "weight_decay": 0.0 }, "mean": 0.0007705322786932811 }, { "cfg": { "lr": 0.006, "weight_decay": 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": 0.000772874424001202, "std": 0.00016036009165282255, "per_seed": [ 0.0006930093513801694, 0.0008406945853494108, 0.0005222903564572334, 0.0009326299768872559, 0.0005455962382256985, 0.0008512879139743745, 0.0008067093440331519, 0.0009907776257023215 ], "n": 8, "best_cfg": { "lr": 0.006, "weight_decay": 0.0001, "gamma": 0.01 }, "sweep": [ { "cfg": { "lr": 0.0015, "weight_decay": 0.0001, "gamma": 0.01 }, "mean": 0.0017785793825169094, "per_seed": [ 0.0009525532950647175, 0.0017182347364723682, 0.0013132664607837796, 0.0025045962538570166, 0.0009531762916594744, 0.001337766065262258, 0.002284065820276737, 0.0031649761367589235 ] }, { "cfg": { "lr": 0.0015, "weight_decay": 0.0001, "gamma": 0.03 }, "mean": 0.001787718734703958, "per_seed": [ 0.0009378251270391047, 0.001730192918330431, 0.001301152864471078, 0.0025130226276814938, 0.0009530456154607236, 0.001333152875304222, 0.002290173200890422, 0.0032431846484541893 ] }, { "cfg": { "lr": 0.0015, "weight_decay": 0.0001, "gamma": 0.1 }, "mean": 0.0018141339023713954, "per_seed": [ 0.0009042767342180014, 0.001764905173331499, 0.0012808581814169884, 0.0025383830070495605, 0.0009604471852071583, 0.0013194901403039694, 0.002280787331983447, 0.003463923465460539 ] }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001, "gamma": 0.01 }, "mean": 0.0010871518170461059, "per_seed": [ 0.0009137207525782287, 0.0008784217061474919, 0.0006754152127541602, 0.0016324227908626199, 0.0006958629237487912, 0.0008044865680858493, 0.0014184740139171481, 0.0016784105682745576 ] }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001, "gamma": 0.03 }, "mean": 0.001080300826288294, "per_seed": [ 0.0009089682716876268, 0.0008737270254641771, 0.0006606929819099605, 0.0016298916889354587, 0.00069375493330881, 0.0008071218035183847, 0.0014060401590541005, 0.001662209746427834 ] }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001, "gamma": 0.1 }, "mean": 0.0010688186739571393, "per_seed": [ 0.0008881261455826461, 0.0008660596213303506, 0.0006356937228702009, 0.001619623857550323, 0.0006789837498217821, 0.0008229009690694511, 0.0013761218870058656, 0.0016630394384264946 ] }, { "cfg": { "lr": 0.006, "weight_decay": 0.0001, "gamma": 0.01 }, "mean": 0.000772874424001202, "per_seed": [ 0.0006930093513801694, 0.0008406945853494108, 0.0005222903564572334, 0.0009326299768872559, 0.0005455962382256985, 0.0008512879139743745, 0.0008067093440331519, 0.0009907776257023215 ] }, { "cfg": { "lr": 0.006, "weight_decay": 0.0001, "gamma": 0.03 }, "mean": 0.0007740218861727044, "per_seed": [ 0.0006858397973701358, 0.0008134624222293496, 0.0005332473665475845, 0.0009384911973029375, 0.000545005255844444, 0.0008184867911040783, 0.0008121238206513226, 0.0010455184383317828 ] }, { "cfg": { "lr": 0.006, "weight_decay": 0.0001, "gamma": 0.1 }, "mean": 0.0007736754487268627, "per_seed": [ 0.0006801989511586726, 0.000721059856005013, 0.0005168036441318691, 0.0009618915501050651, 0.0006032949313521385, 0.0007013163412921131, 0.000832107849419117, 0.001172730466350913 ] } ] }, "comparison": { "delta_mean": 1.6353733371943235e-06, "idea_wins": 4, "n_pairs": 8, "per_seed_diffs": [ -4.104978870600462e-06, -1.2010917998850346e-05, 8.269678801298141e-06, 3.1457748264074326e-06, -2.409098669886589e-06, -1.0057294275611639e-05, 1.796637661755085e-06, 2.8453185223042965e-05 ], "p_value": 0.79715, "mde": 1.066015482771443e-05, "mde_rel_pct": 1.382211497010744, "verdict": "no measurable effect", "system_worked": false }, "mechanism_signature": { "track_match": "dynamics: controlled pendulum rollout; odd sign-reversal channel", "prediction": "post-probe control suppresses symmetry-breaking mode, so late A2 proxy falls below probe A2 proxy", "predicted_vs_observed": { "predicted_late_to_probe_ratio": "< 1.0", "observed_mean_ratio": 0.16168932957134385, "per_seed": [ { "seed": 0, "baseline_A2_proxy": 0.0066139898262917995, "idea_probe_A2_proxy": 0.028515702579170465, "idea_late_A2_proxy": 0.007312557194381952 }, { "seed": 1, "baseline_A2_proxy": 0.00954736303538084, "idea_probe_A2_proxy": 0.06434168418248494, "idea_late_A2_proxy": 0.013520444432894388 }, { "seed": 2, "baseline_A2_proxy": 0.009730289690196514, "idea_probe_A2_proxy": 0.12297303229570389, "idea_late_A2_proxy": 0.011106709018349648 }, { "seed": 3, "baseline_A2_proxy": 0.005885358899831772, "idea_probe_A2_proxy": 0.02378938440233469, "idea_late_A2_proxy": 0.006580424960702658 }, { "seed": 4, "baseline_A2_proxy": 0.004799993708729744, "idea_probe_A2_proxy": 0.03924484054247538, "idea_late_A2_proxy": 0.0038459034791837135 }, { "seed": 5, "baseline_A2_proxy": 0.014292827807366848, "idea_probe_A2_proxy": 0.0702164676040411, "idea_late_A2_proxy": 0.00915219308808446 }, { "seed": 6, "baseline_A2_proxy": 0.0031815164256840944, "idea_probe_A2_proxy": 0.11305895199378331, "idea_late_A2_proxy": 0.006548060104250908 }, { "seed": 7, "baseline_A2_proxy": 0.013455699197947979, "idea_probe_A2_proxy": 0.07279480372865994, "idea_late_A2_proxy": 0.012648270465433598 } ] }, "confirmed": true } }