{ "bench_version": 1, "track": "sequence", "model": "transformer_tiny", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.001 }, "sweep": [ { "cfg": { "lr": 0.001 }, "mean": 0.5707684904336929 }, { "cfg": { "lr": 0.003 }, "mean": 0.6878322213888168 }, { "cfg": { "lr": 0.006 }, "mean": 1.0211272835731506 } ], "full": { "mean": 0.5598551481962204, "std": 0.02523705019158861, "per_seed": [ 0.5729890465736389, 0.5925205945968628, 0.5500922203063965, 0.5674721002578735, 0.5691339373588562, 0.5163435935974121, 0.5841162204742432, 0.52617347240448 ], "n": 8 } }, "idea": { "mean": 0.5598490461707115, "std": 0.025273703128400025, "per_seed": [ 0.5730356574058533, 0.5925182104110718, 0.5501575469970703, 0.5674620866775513, 0.5691909790039062, 0.516342043876648, 0.5841178894042969, 0.5259679555892944 ], "n": 8 }, "comparison": { "delta_mean": -6.102025508880615e-06, "idea_wins": 4, "n_pairs": 8, "per_seed_diffs": [ 4.661083221435547e-05, -2.384185791015625e-06, 6.532669067382812e-05, -1.0013580322265625e-05, 5.704164505004883e-05, -1.5497207641601562e-06, 1.6689300537109375e-06, -0.00020551681518554688 ], "p_value": 0.8987, "mde": 7.188999984046296e-05, "mde_rel_pct": 0.012840821428914797, "verdict": "no measurable effect", "system_worked": false }, "mechanism_signature": { "method": "MI-gradient inverse variance at 64-d token latent", "best_idea_cfg": { "lr": 0.001 }, "idea_sweep": [ { "cfg": { "lr": 0.001 }, "result": { "mean": 0.5598490461707115, "std": 0.025273703128400025, "per_seed": [ 0.5730356574058533, 0.5925182104110718, 0.5501575469970703, 0.5674620866775513, 0.5691909790039062, 0.516342043876648, 0.5841178894042969, 0.5259679555892944 ], "n": 8 } }, { "cfg": { "lr": 0.003 }, "result": { "mean": 0.6711730286478996, "std": 0.11012775840681698, "per_seed": [ 0.5534999966621399, 0.57029789686203, 0.8852774500846863, 0.741381824016571, 0.5719724297523499, 0.75211101770401, 0.6959620714187622, 0.5988815426826477 ], "n": 8 } }, { "cfg": { "lr": 0.006 }, "result": { "mean": 1.010021261870861, "std": 0.1928573253470958, "per_seed": [ 0.9215983152389526, 0.815628170967102, 0.754155158996582, 0.9170165657997131, 1.3768141269683838, 0.983137845993042, 1.1562342643737793, 1.1555856466293335 ], "n": 8 } } ], "predicted_vs_observed": { "metric": 0.5730356574058533, "q": [ 1.0012232065200806, 0.999777615070343, 0.9991872906684875, 1.001699447631836, 1.0009998083114624, 0.9999410510063171, 0.9996676445007324, 1.0007892847061157, 0.9993820190429688, 1.0001580715179443, 0.9999819993972778, 0.9993095397949219, 0.9994521141052246, 0.9997609853744507, 0.9999178647994995, 0.9990708827972412, 0.9990319609642029, 0.9996174573898315, 1.0006672143936157, 1.000030279159546, 1.000243902206421, 0.9987349510192871, 1.0009965896606445, 0.9990262389183044, 1.001396894454956, 0.9992781281471252, 0.99986332654953, 1.0000636577606201, 1.0003782510757446, 0.999755322933197, 0.9989909529685974, 1.0014028549194336, 0.9997116923332214, 1.0010855197906494, 1.0008490085601807, 0.9986003637313843, 0.9999604225158691, 1.0011999607086182, 0.9982680082321167, 0.9994734525680542, 0.9993472695350647, 0.9998814463615417, 0.9986863732337952, 1.000624656677246, 0.9999277591705322, 0.9998847246170044, 1.000219464302063, 0.9992424845695496, 0.9991764426231384, 0.9996323585510254, 0.999434769153595, 1.0023554563522339, 1.0001237392425537, 1.0017119646072388, 0.9994488954544067, 1.0003513097763062, 1.0015864372253418, 0.9987058639526367, 1.0016520023345947, 1.0003422498703003, 0.9981855750083923, 1.000388741493225, 0.999549150466919, 1.0005910396575928 ], "variances": [ 0.19975546002388, 0.20004430413246155, 0.20016248524188995, 0.1996604949235916, 0.199800044298172, 0.20001161098480225, 0.20006629824638367, 0.19984206557273865, 0.20012347400188446, 0.19996818900108337, 0.20000341534614563, 0.20013800263404846, 0.20010945200920105, 0.20004761219024658, 0.20001623034477234, 0.20018580555915833, 0.2001935988664627, 0.20007634162902832, 0.19986644387245178, 0.19999375939369202, 0.19995103776454926, 0.20025312900543213, 0.19980069994926453, 0.20019474625587463, 0.1997208297252655, 0.20014429092407227, 0.20002715289592743, 0.19998708367347717, 0.19992418587207794, 0.2000487595796585, 0.2002018243074417, 0.19971963763237, 0.2000574767589569, 0.19978293776512146, 0.19983014464378357, 0.2002801150083542, 0.20000772178173065, 0.19976010918617249, 0.20034679770469666, 0.20010516047477722, 0.200130432844162, 0.20002353191375732, 0.2002628743648529, 0.19987496733665466, 0.20001426339149475, 0.2000228613615036, 0.19995592534542084, 0.20015141367912292, 0.20016464591026306, 0.20007336139678955, 0.20011290907859802, 0.19952982664108276, 0.19997505843639374, 0.19965800642967224, 0.20011010766029358, 0.19992956519126892, 0.1996830254793167, 0.20025897026062012, 0.1996699571609497, 0.19993138313293457, 0.20036335289478302, 0.1999220997095108, 0.20009002089500427, 0.1998816728591919 ], "predicted_latent_noise_power": 12.799999237060547, "observed_output_perturbation_mse": 0.06943081552162766, "observed_output_perturbation_std": 0.006431090703341278, "predicted_vs_observed_ratio": 0.00542428278593961, "confirmed": true } }, "paired_seed_protocol": { "seeds": [ 0, 1, 2, 3, 4, 5, 6, 7 ], "epochs": 8, "batch": 128, "average_variance": 0.2 } }