{ "bench_version": 1, "track": "dynamics", "model": "rnn_small_step_gru", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.005, "mode": "iid" }, "sweep": [ { "cfg": { "lr": 0.001, "mode": "iid" }, "mean": 0.055274722166359425 }, { "cfg": { "lr": 0.002, "mode": "iid" }, "mean": 0.03774398844689131 }, { "cfg": { "lr": 0.003, "mode": "iid" }, "mean": 0.03395134303718805 }, { "cfg": { "lr": 0.005, "mode": "iid" }, "mean": 0.027184664737433195 } ], "full": { "mean": 0.027283388655632734, "std": 0.002202312165538543, "per_seed": [ 0.026371225714683533, 0.029238859191536903, 0.024704264476895332, 0.028424309566617012, 0.02377459779381752, 0.028389407321810722, 0.026591340079903603, 0.03077310509979725 ], "n": 8 }, "idea_grid": [ { "cfg": { "lr": 0.005, "gamma": 0.3 }, "result": { "mean": 0.025829498656094074, "std": 0.004832996344710331, "per_seed": [ 0.024646440520882607, 0.03045663796365261, 0.021562719717621803, 0.035005223006010056, 0.02129276655614376, 0.025303291156888008, 0.019978370517492294, 0.028390539810061455 ], "n": 8 }, "signature_samples": [ { "seed": 0, "metric": 0.024646440520882607, "hidden_corr_lags": [ 0.6630269404071175, 0.5190174493427347, 0.4422342212494631, 0.3680763148053659, 0.3278020144627789, 0.2862978241753326, 0.2412090288981012 ] }, { "seed": 1, "metric": 0.03045663796365261, "hidden_corr_lags": [ 0.6558773935907547, 0.5223715348082428, 0.43886990260874603, 0.37703245036602556, 0.3213551689286596, 0.2888605860363376, 0.2578390283392032 ] }, { "seed": 2, "metric": 0.021562719717621803, "hidden_corr_lags": [ 0.6941417197981021, 0.5650620747828666, 0.4855752673149188, 0.4213088437221936, 0.3834199477681708, 0.3304678752637303, 0.28157898951072124 ] }, { "seed": 3, "metric": 0.035005223006010056, "hidden_corr_lags": [ 0.6570608148081333, 0.5009651676177592, 0.40564033635438773, 0.32656343606835014, 0.27598215558039724, 0.22561703828793092, 0.19726381926073394 ] }, { "seed": 4, "metric": 0.02129276655614376, "hidden_corr_lags": [ 0.6826800603773961, 0.5467952964304799, 0.4468693866002476, 0.3782948169847446, 0.3253786488327958, 0.29587949094718413, 0.2301728069470637 ] }, { "seed": 5, "metric": 0.025303291156888008, "hidden_corr_lags": [ 0.708347564776786, 0.5759075676091028, 0.5077637712155896, 0.448388369326119, 0.3987969864771873, 0.33574424929679675, 0.28392445027096763 ] }, { "seed": 6, "metric": 0.019978370517492294, "hidden_corr_lags": [ 0.673128447135754, 0.5402513645806349, 0.45873630730381026, 0.3854459282621681, 0.32234771024755, 0.2952321479991386, 0.2492059622628454 ] }, { "seed": 7, "metric": 0.028390539810061455, "hidden_corr_lags": [ 0.7141917699225196, 0.5837811124829955, 0.4948940626715219, 0.4254349838699835, 0.3579367016144324, 0.30836113208571736, 0.29892915453046004 ] } ] }, { "cfg": { "lr": 0.002, "gamma": 0.75 }, "result": { "mean": 0.03631573240272701, "std": 0.004775914433401002, "per_seed": [ 0.0330687016248703, 0.0394829586148262, 0.030048971995711327, 0.04279492795467377, 0.031434860080480576, 0.03951079398393631, 0.0323273167014122, 0.04185732826590538 ], "n": 8 }, "signature_samples": [ { "seed": 0, "metric": 0.0330687016248703, "hidden_corr_lags": [ 0.6102964978441251, 0.42560961106117584, 0.31946573637749986, 0.23220161620796825, 0.19459308793457936, 0.15619064232072974, 0.1421222028155319 ] }, { "seed": 1, "metric": 0.0394829586148262, "hidden_corr_lags": [ 0.6005485181378184, 0.41083370762196114, 0.303425006453075, 0.2291316727228182, 0.16831260553042063, 0.14304766995065002, 0.1389789160825483 ] }, { "seed": 2, "metric": 0.030048971995711327, "hidden_corr_lags": [ 0.622951561402403, 0.4462168785833603, 0.3393174411120082, 0.2702143518879489, 0.22021449688594993, 0.17974238665253822, 0.12501642329757462 ] }, { "seed": 3, "metric": 0.04279492795467377, "hidden_corr_lags": [ 0.619714269053892, 0.4200118264269483, 0.2923894487354256, 0.20448650296559157, 0.15068702526393485, 0.11022987153793977, 0.09068743549906201 ] }, { "seed": 4, "metric": 0.031434860080480576, "hidden_corr_lags": [ 0.6220310045067547, 0.4415479430121763, 0.30980329235113746, 0.2292403308903853, 0.17705759851987635, 0.15192179639746586, 0.10008725515824768 ] }, { "seed": 5, "metric": 0.03951079398393631, "hidden_corr_lags": [ 0.6492663473855156, 0.4752660395260816, 0.37186456254988143, 0.2993911429379207, 0.2506879111917356, 0.1706674850386674, 0.12579496596449938 ] }, { "seed": 6, "metric": 0.0323273167014122, "hidden_corr_lags": [ 0.6249857539507511, 0.43661329470398996, 0.31667409773638083, 0.23511431262697952, 0.17360203585510836, 0.1128293631726479, 0.08316472431468136 ] }, { "seed": 7, "metric": 0.04185732826590538, "hidden_corr_lags": [ 0.6477746076383196, 0.4837374633747622, 0.3739317574253478, 0.2873661106042647, 0.20245513007710744, 0.16273499077256248, 0.15616598085300054 ] } ] }, { "cfg": { "lr": 0.005, "gamma": 1.25 }, "result": { "mean": 0.029610266210511327, "std": 0.004106556225593385, "per_seed": [ 0.02844509482383728, 0.03574374318122864, 0.02142076939344406, 0.033146582543849945, 0.02610703371465206, 0.029993567615747452, 0.030954143032431602, 0.031071195378899574 ], "n": 8 }, "signature_samples": [ { "seed": 0, "metric": 0.02844509482383728, "hidden_corr_lags": [ 0.5661622634875924, 0.35935370448605075, 0.24470705832316364, 0.15001805109697283, 0.1170011517966459, 0.09007776492377811, 0.09657207494809344 ] }, { "seed": 1, "metric": 0.03574374318122864, "hidden_corr_lags": [ 0.5424371578116096, 0.3316706199975801, 0.22097801235442388, 0.15025007256407527, 0.09818635242882416, 0.07305968834285463, 0.07959775154561222 ] }, { "seed": 2, "metric": 0.02142076939344406, "hidden_corr_lags": [ 0.5498584303173741, 0.33527281959180183, 0.22394417898051278, 0.16902627448160581, 0.10906020598843227, 0.06903056036991773, 0.025547922591465724 ] }, { "seed": 3, "metric": 0.033146582543849945, "hidden_corr_lags": [ 0.5617111930595085, 0.33975738810671835, 0.20733083952696452, 0.13007694009658544, 0.07507689865068767, 0.04378026547839316, 0.03422802431216209 ] }, { "seed": 4, "metric": 0.02610703371465206, "hidden_corr_lags": [ 0.5815045217313184, 0.3783075053868034, 0.24312359573973874, 0.16568378259297578, 0.1238111278061614, 0.098778516460376, 0.03670061476829134 ] }, { "seed": 5, "metric": 0.029993567615747452, "hidden_corr_lags": [ 0.581782723753792, 0.3807485682563384, 0.25706393866837807, 0.1911667666655277, 0.15547337653618265, 0.09331923344449601, 0.05540279009505447 ] }, { "seed": 6, "metric": 0.030954143032431602, "hidden_corr_lags": [ 0.5751503122969772, 0.3557994072626178, 0.22949450651436587, 0.150793315022976, 0.10195281481195258, 0.056958478116733595, 0.05140981554943419 ] }, { "seed": 7, "metric": 0.031071195378899574, "hidden_corr_lags": [ 0.5817921571206278, 0.38842194382538286, 0.26982850115764956, 0.1773631426455767, 0.10070742394505952, 0.06197348489515356, 0.06083095421292512 ] } ] } ] }, "idea": { "mean": 0.025829498656094074, "std": 0.004832996344710331, "per_seed": [ 0.024646440520882607, 0.03045663796365261, 0.021562719717621803, 0.035005223006010056, 0.02129276655614376, 0.025303291156888008, 0.019978370517492294, 0.028390539810061455 ], "n": 8 }, "comparison": { "delta_mean": -0.00145388999953866, "idea_wins": 6, "n_pairs": 8, "per_seed_diffs": [ -0.0017247851938009262, 0.0012177787721157074, -0.003141544759273529, 0.0065809134393930435, -0.0024818312376737595, -0.0030861161649227142, -0.006612969562411308, -0.002382565289735794 ], "p_value": 0.31345, "mde": 0.0032509812188714292, "mde_rel_pct": 11.915606451621084, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "mechanism_signature": { "quantity": "trained hidden-state cross-step correlation on dynamics test trajectories", "gamma": 0.3, "predicted_loglog_slope": -0.3, "observed_loglog_slope": -0.7032384334932155, "mean_abs_corr_by_lag": [ 0.6810568388520706, 0.5442689459568522, 0.4600729069148357, 0.39131814292561884, 0.3391274167389965, 0.295807543011521, 0.255015405002512 ], "confirmed": true }, "baseline_architecture": "8 independent GRUCell steps", "idea_architecture": "same 8 GRUCell steps, coordinatewise power-law covariance" }, "idea_sweep": [ { "cfg": { "lr": 0.005, "gamma": 0.3 }, "result": { "mean": 0.025829498656094074, "std": 0.004832996344710331, "per_seed": [ 0.024646440520882607, 0.03045663796365261, 0.021562719717621803, 0.035005223006010056, 0.02129276655614376, 0.025303291156888008, 0.019978370517492294, 0.028390539810061455 ], "n": 8 }, "signature_samples": [ { "seed": 0, "metric": 0.024646440520882607, "hidden_corr_lags": [ 0.6630269404071175, 0.5190174493427347, 0.4422342212494631, 0.3680763148053659, 0.3278020144627789, 0.2862978241753326, 0.2412090288981012 ] }, { "seed": 1, "metric": 0.03045663796365261, "hidden_corr_lags": [ 0.6558773935907547, 0.5223715348082428, 0.43886990260874603, 0.37703245036602556, 0.3213551689286596, 0.2888605860363376, 0.2578390283392032 ] }, { "seed": 2, "metric": 0.021562719717621803, "hidden_corr_lags": [ 0.6941417197981021, 0.5650620747828666, 0.4855752673149188, 0.4213088437221936, 0.3834199477681708, 0.3304678752637303, 0.28157898951072124 ] }, { "seed": 3, "metric": 0.035005223006010056, "hidden_corr_lags": [ 0.6570608148081333, 0.5009651676177592, 0.40564033635438773, 0.32656343606835014, 0.27598215558039724, 0.22561703828793092, 0.19726381926073394 ] }, { "seed": 4, "metric": 0.02129276655614376, "hidden_corr_lags": [ 0.6826800603773961, 0.5467952964304799, 0.4468693866002476, 0.3782948169847446, 0.3253786488327958, 0.29587949094718413, 0.2301728069470637 ] }, { "seed": 5, "metric": 0.025303291156888008, "hidden_corr_lags": [ 0.708347564776786, 0.5759075676091028, 0.5077637712155896, 0.448388369326119, 0.3987969864771873, 0.33574424929679675, 0.28392445027096763 ] }, { "seed": 6, "metric": 0.019978370517492294, "hidden_corr_lags": [ 0.673128447135754, 0.5402513645806349, 0.45873630730381026, 0.3854459282621681, 0.32234771024755, 0.2952321479991386, 0.2492059622628454 ] }, { "seed": 7, "metric": 0.028390539810061455, "hidden_corr_lags": [ 0.7141917699225196, 0.5837811124829955, 0.4948940626715219, 0.4254349838699835, 0.3579367016144324, 0.30836113208571736, 0.29892915453046004 ] } ] }, { "cfg": { "lr": 0.002, "gamma": 0.75 }, "result": { "mean": 0.03631573240272701, "std": 0.004775914433401002, "per_seed": [ 0.0330687016248703, 0.0394829586148262, 0.030048971995711327, 0.04279492795467377, 0.031434860080480576, 0.03951079398393631, 0.0323273167014122, 0.04185732826590538 ], "n": 8 }, "signature_samples": [ { "seed": 0, "metric": 0.0330687016248703, "hidden_corr_lags": [ 0.6102964978441251, 0.42560961106117584, 0.31946573637749986, 0.23220161620796825, 0.19459308793457936, 0.15619064232072974, 0.1421222028155319 ] }, { "seed": 1, "metric": 0.0394829586148262, "hidden_corr_lags": [ 0.6005485181378184, 0.41083370762196114, 0.303425006453075, 0.2291316727228182, 0.16831260553042063, 0.14304766995065002, 0.1389789160825483 ] }, { "seed": 2, "metric": 0.030048971995711327, "hidden_corr_lags": [ 0.622951561402403, 0.4462168785833603, 0.3393174411120082, 0.2702143518879489, 0.22021449688594993, 0.17974238665253822, 0.12501642329757462 ] }, { "seed": 3, "metric": 0.04279492795467377, "hidden_corr_lags": [ 0.619714269053892, 0.4200118264269483, 0.2923894487354256, 0.20448650296559157, 0.15068702526393485, 0.11022987153793977, 0.09068743549906201 ] }, { "seed": 4, "metric": 0.031434860080480576, "hidden_corr_lags": [ 0.6220310045067547, 0.4415479430121763, 0.30980329235113746, 0.2292403308903853, 0.17705759851987635, 0.15192179639746586, 0.10008725515824768 ] }, { "seed": 5, "metric": 0.03951079398393631, "hidden_corr_lags": [ 0.6492663473855156, 0.4752660395260816, 0.37186456254988143, 0.2993911429379207, 0.2506879111917356, 0.1706674850386674, 0.12579496596449938 ] }, { "seed": 6, "metric": 0.0323273167014122, "hidden_corr_lags": [ 0.6249857539507511, 0.43661329470398996, 0.31667409773638083, 0.23511431262697952, 0.17360203585510836, 0.1128293631726479, 0.08316472431468136 ] }, { "seed": 7, "metric": 0.04185732826590538, "hidden_corr_lags": [ 0.6477746076383196, 0.4837374633747622, 0.3739317574253478, 0.2873661106042647, 0.20245513007710744, 0.16273499077256248, 0.15616598085300054 ] } ] }, { "cfg": { "lr": 0.005, "gamma": 1.25 }, "result": { "mean": 0.029610266210511327, "std": 0.004106556225593385, "per_seed": [ 0.02844509482383728, 0.03574374318122864, 0.02142076939344406, 0.033146582543849945, 0.02610703371465206, 0.029993567615747452, 0.030954143032431602, 0.031071195378899574 ], "n": 8 }, "signature_samples": [ { "seed": 0, "metric": 0.02844509482383728, "hidden_corr_lags": [ 0.5661622634875924, 0.35935370448605075, 0.24470705832316364, 0.15001805109697283, 0.1170011517966459, 0.09007776492377811, 0.09657207494809344 ] }, { "seed": 1, "metric": 0.03574374318122864, "hidden_corr_lags": [ 0.5424371578116096, 0.3316706199975801, 0.22097801235442388, 0.15025007256407527, 0.09818635242882416, 0.07305968834285463, 0.07959775154561222 ] }, { "seed": 2, "metric": 0.02142076939344406, "hidden_corr_lags": [ 0.5498584303173741, 0.33527281959180183, 0.22394417898051278, 0.16902627448160581, 0.10906020598843227, 0.06903056036991773, 0.025547922591465724 ] }, { "seed": 3, "metric": 0.033146582543849945, "hidden_corr_lags": [ 0.5617111930595085, 0.33975738810671835, 0.20733083952696452, 0.13007694009658544, 0.07507689865068767, 0.04378026547839316, 0.03422802431216209 ] }, { "seed": 4, "metric": 0.02610703371465206, "hidden_corr_lags": [ 0.5815045217313184, 0.3783075053868034, 0.24312359573973874, 0.16568378259297578, 0.1238111278061614, 0.098778516460376, 0.03670061476829134 ] }, { "seed": 5, "metric": 0.029993567615747452, "hidden_corr_lags": [ 0.581782723753792, 0.3807485682563384, 0.25706393866837807, 0.1911667666655277, 0.15547337653618265, 0.09331923344449601, 0.05540279009505447 ] }, { "seed": 6, "metric": 0.030954143032431602, "hidden_corr_lags": [ 0.5751503122969772, 0.3557994072626178, 0.22949450651436587, 0.150793315022976, 0.10195281481195258, 0.056958478116733595, 0.05140981554943419 ] }, { "seed": 7, "metric": 0.031071195378899574, "hidden_corr_lags": [ 0.5817921571206278, 0.38842194382538286, 0.26982850115764956, 0.1773631426455767, 0.10070742394505952, 0.06197348489515356, 0.06083095421292512 ] } ] } ], "protocol_note": "Dynamics chosen because the idea concerns recurrent depth and stability; 8 paired seeds, baseline LR sweep, idea 3-config sweep." }