{ "bench_version": 1, "track": "sequence", "model": "transformer_tiny", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.0015, "weight_decay": 0.0 }, "sweep": [ { "cfg": { "lr": 0.0015, "weight_decay": 0.0 }, "mean": 0.5785842016339302 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0 }, "mean": 0.6899479925632477 }, { "cfg": { "lr": 0.006, "weight_decay": 0.0 }, "mean": 0.8105297237634659 } ], "full": { "mean": 0.562101598829031, "std": 0.08060035643973527, "per_seed": [ 0.6448022127151489, 0.4093892276287079, 0.5553309321403503, 0.7048144340515137, 0.5405545234680176, 0.5604735016822815, 0.5436524748802185, 0.5377954840660095 ], "n": 8 } }, "idea": { "mean": 0.5841716527938843, "std": 0.05076107251588834, "per_seed": [ 0.6083592772483826, 0.5841273665428162, 0.6231781840324402, 0.5197466015815735, 0.6223173141479492, 0.5627732872962952, 0.654932975769043, 0.49793821573257446 ], "n": 8 }, "comparison": { "delta_mean": 0.022070053964853287, "idea_wins": 3, "n_pairs": 8, "per_seed_diffs": [ -0.03644293546676636, 0.17473813891410828, 0.06784725189208984, -0.18506783246994019, 0.08176279067993164, 0.002299785614013672, 0.11128050088882446, -0.03985726833343506 ], "p_value": 0.5933, "mde": 0.09347444437257611, "mde_rel_pct": 16.629457124352946, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "mechanism_signature": { "prediction": "identity insertion and O(eta) untied response", "confirmed": true, "idea_best_cfg": { "lr": 0.0015, "weight_decay": 0.0 }, "per_seed": [ { "seed": 0, "insertion_relative_output_change": 1.742963640083417e-08, "trained_pair_residual": 0.013553540222346783, "untied_eta_response": [ [ 1e-05, 6.882298180244106e-07 ], [ 0.0001, 6.846436008345336e-06 ], [ 0.001, 6.854822277091444e-05 ] ], "trained_loglog_slope": 0.9991313573267615 }, { "seed": 1, "insertion_relative_output_change": 1.768672674984373e-08, "trained_pair_residual": 0.01374770887196064, "untied_eta_response": [ [ 1e-05, 1.540886728434998e-06 ], [ 0.0001, 1.5476085536647588e-05 ], [ 0.001, 0.0001548135041957721 ] ], "trained_loglog_slope": 1.0010190631783853 }, { "seed": 2, "insertion_relative_output_change": 1.7788378769978408e-08, "trained_pair_residual": 0.01247485727071762, "untied_eta_response": [ [ 1e-05, 2.696924639167264e-06 ], [ 0.0001, 2.6614063244778663e-05 ], [ 0.001, 0.0002659428573679179 ] ], "trained_loglog_slope": 0.9969597598038572 }, { "seed": 3, "insertion_relative_output_change": 1.678153260797899e-08, "trained_pair_residual": 0.012198672629892826, "untied_eta_response": [ [ 1e-05, 3.872593197229435e-07 ], [ 0.0001, 3.2745517728471896e-06 ], [ 0.001, 3.2737443689256907e-05 ] ], "trained_loglog_slope": 0.9635214430977415 }, { "seed": 4, "insertion_relative_output_change": 1.788907688649033e-08, "trained_pair_residual": 0.012925044633448124, "untied_eta_response": [ [ 1e-05, 1.7712729913910152e-06 ], [ 0.0001, 1.768083166098222e-05 ], [ 0.001, 0.00017684148042462766 ] ], "trained_loglog_slope": 0.9996493207179143 }, { "seed": 5, "insertion_relative_output_change": 1.6752988329926666e-08, "trained_pair_residual": 0.011602357029914856, "untied_eta_response": [ [ 1e-05, 9.838264531936147e-07 ], [ 0.0001, 9.362133823742624e-06 ], [ 0.001, 9.372592467116192e-05 ] ], "trained_loglog_slope": 0.9894706190051281 }, { "seed": 6, "insertion_relative_output_change": 1.6823253901065982e-08, "trained_pair_residual": 0.012010437436401844, "untied_eta_response": [ [ 1e-05, 1.4330839803733397e-06 ], [ 0.0001, 1.4343587281473447e-05 ], [ 0.001, 0.00014369039854500443 ] ], "trained_loglog_slope": 1.0005780540284896 }, { "seed": 7, "insertion_relative_output_change": 1.7179964117985946e-08, "trained_pair_residual": 0.013055686838924885, "untied_eta_response": [ [ 1e-05, 6.872157314319338e-07 ], [ 0.0001, 6.427855623769574e-06 ], [ 0.001, 6.434921670006588e-05 ] ], "trained_loglog_slope": 0.9857250860758257 }, { "seed": 0, "insertion_relative_output_change": 1.742963640083417e-08, "trained_pair_residual": 0.024646366015076637, "untied_eta_response": [ [ 1e-05, 6.151723823677457e-07 ], [ 0.0001, 6.120769739936804e-06 ], [ 0.001, 6.104174099164084e-05 ] ], "trained_loglog_slope": 0.9983150409119986 }, { "seed": 1, "insertion_relative_output_change": 1.768672674984373e-08, "trained_pair_residual": 0.026445627212524414, "untied_eta_response": [ [ 1e-05, 1.0641194876370719e-06 ], [ 0.0001, 1.0895923878706526e-05 ], [ 0.001, 0.00010886218660743907 ] ], "trained_loglog_slope": 1.0049433283222116 }, { "seed": 2, "insertion_relative_output_change": 1.7788378769978408e-08, "trained_pair_residual": 0.023685336112976074, "untied_eta_response": [ [ 1e-05, 1.0162499393118196e-06 ], [ 0.0001, 1.0251549610984512e-05 ], [ 0.001, 0.00010272490908391774 ] ], "trained_loglog_slope": 1.0023376164486255 }, { "seed": 3, "insertion_relative_output_change": 1.678153260797899e-08, "trained_pair_residual": 0.024265628308057785, "untied_eta_response": [ [ 1e-05, 2.682033937162487e-06 ], [ 0.0001, 2.6792338758241385e-05 ], [ 0.001, 0.0002678318414837122 ] ], "trained_loglog_slope": 0.9996989692102766 }, { "seed": 4, "insertion_relative_output_change": 1.788907688649033e-08, "trained_pair_residual": 0.02498849853873253, "untied_eta_response": [ [ 1e-05, 1.6889165408429108e-06 ], [ 0.0001, 1.6884097931324504e-05 ], [ 0.001, 0.00016868698003236204 ] ], "trained_loglog_slope": 0.9997366870611765 }, { "seed": 5, "insertion_relative_output_change": 1.6752988329926666e-08, "trained_pair_residual": 0.02313663437962532, "untied_eta_response": [ [ 1e-05, 1.1504967005748767e-06 ], [ 0.0001, 1.1409663784434088e-05 ], [ 0.001, 0.00011402922245906666 ] ], "trained_loglog_slope": 0.9980653927231617 }, { "seed": 6, "insertion_relative_output_change": 1.6823253901065982e-08, "trained_pair_residual": 0.023537755012512207, "untied_eta_response": [ [ 1e-05, 1.739130652822496e-06 ], [ 0.0001, 1.758110556693282e-05 ], [ 0.001, 0.00017618776473682374 ] ], "trained_loglog_slope": 1.0028217680103706 }, { "seed": 7, "insertion_relative_output_change": 1.7179964117985946e-08, "trained_pair_residual": 0.024072429165244102, "untied_eta_response": [ [ 1e-05, 7.772475214551378e-07 ], [ 0.0001, 7.599023319926346e-06 ], [ 0.001, 7.574163464596495e-05 ] ], "trained_loglog_slope": 0.9943876639826758 }, { "seed": 0, "insertion_relative_output_change": 1.742963640083417e-08, "trained_pair_residual": 0.049147915095090866, "untied_eta_response": [ [ 1e-05, 1.9303267890791176e-06 ], [ 0.0001, 1.8857945178751834e-05 ], [ 0.001, 0.0001881620701169595 ] ], "trained_loglog_slope": 0.9944506222877422 }, { "seed": 1, "insertion_relative_output_change": 1.768672674984373e-08, "trained_pair_residual": 0.04954760521650314, "untied_eta_response": [ [ 1e-05, 8.311401870741975e-06 ], [ 0.0001, 8.277293818537146e-05 ], [ 0.001, 0.0008280269103124738 ] ], "trained_loglog_slope": 0.9991850847890872 }, { "seed": 2, "insertion_relative_output_change": 1.7788378769978408e-08, "trained_pair_residual": 0.04762578755617142, "untied_eta_response": [ [ 1e-05, 7.512046067859046e-06 ], [ 0.0001, 7.488950359402224e-05 ], [ 0.001, 0.0007490166462957859 ] ], "trained_loglog_slope": 0.9993666135316805 }, { "seed": 3, "insertion_relative_output_change": 1.678153260797899e-08, "trained_pair_residual": 0.045538078993558884, "untied_eta_response": [ [ 1e-05, 1.7467951693106443e-05 ], [ 0.0001, 0.000174284607055597 ], [ 0.001, 0.0017429785802960396 ] ], "trained_loglog_slope": 0.9995250339052046 }, { "seed": 4, "insertion_relative_output_change": 1.788907688649033e-08, "trained_pair_residual": 0.056724853813648224, "untied_eta_response": [ [ 1e-05, 4.201423053018516e-06 ], [ 0.0001, 4.1662235162220895e-05 ], [ 0.001, 0.00041710384539328516 ] ], "trained_loglog_slope": 0.9984238898394366 }, { "seed": 5, "insertion_relative_output_change": 1.6752988329926666e-08, "trained_pair_residual": 0.046788010746240616, "untied_eta_response": [ [ 1e-05, 5.517295448953519e-06 ], [ 0.0001, 5.510038317879662e-05 ], [ 0.001, 0.0005517865065485239 ] ], "trained_loglog_slope": 1.000022417554908 }, { "seed": 6, "insertion_relative_output_change": 1.6823253901065982e-08, "trained_pair_residual": 0.04568944126367569, "untied_eta_response": [ [ 1e-05, 5.431155841506552e-06 ], [ 0.0001, 5.4161326261237264e-05 ], [ 0.001, 0.0005412694299593568 ] ], "trained_loglog_slope": 0.9992606174539198 }, { "seed": 7, "insertion_relative_output_change": 1.7179964117985946e-08, "trained_pair_residual": 0.04901719093322754, "untied_eta_response": [ [ 1e-05, 5.4369570534618106e-06 ], [ 0.0001, 5.419322042143904e-05 ], [ 0.001, 0.0005417013308033347 ] ], "trained_loglog_slope": 0.9992020000011609 } ], "mean_insertion_relative_output_change": 1.7291444720513027e-08, "mean_trained_pair_residual": 0.028601060970686376 } } }