{ "bench_version": 1, "track": "correlated_token_moe_regression", "model": "custom_moe_shared", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.006, "eta": 0.03 }, "sweep": [ { "cfg": { "lr": 0.001, "eta": 0.03 }, "mean": 0.06951743829995394 }, { "cfg": { "lr": 0.001, "eta": 0.1 }, "mean": 0.06951743829995394 }, { "cfg": { "lr": 0.001, "eta": 0.3 }, "mean": 0.06951743829995394 }, { "cfg": { "lr": 0.003, "eta": 0.03 }, "mean": 0.043228451162576675 }, { "cfg": { "lr": 0.003, "eta": 0.1 }, "mean": 0.043228451162576675 }, { "cfg": { "lr": 0.003, "eta": 0.3 }, "mean": 0.043228451162576675 }, { "cfg": { "lr": 0.006, "eta": 0.03 }, "mean": 0.04188114870339632 }, { "cfg": { "lr": 0.006, "eta": 0.1 }, "mean": 0.04188114870339632 }, { "cfg": { "lr": 0.006, "eta": 0.3 }, "mean": 0.04188114870339632 } ], "full": { "mean": 0.04093059757724404, "std": 0.0038662018493219787, "per_seed": [ 0.036128461360931396, 0.04280463978648186, 0.04104797914624214, 0.047543514519929886, 0.038935478776693344, 0.04488188028335571, 0.04060979187488556, 0.03549303486943245 ], "n": 8 } }, "idea": { "mean": 0.04056028323248029, "std": 0.0035352381007606546, "per_seed": [ 0.03617550805211067, 0.042327430099248886, 0.040069837123155594, 0.04594426974654198, 0.039504799991846085, 0.04505091905593872, 0.04003051668405533, 0.03537898510694504 ], "n": 8 }, "comparison": { "delta_mean": -0.0003703143447637558, "idea_wins": 5, "n_pairs": 8, "per_seed_diffs": [ 4.704669117927551e-05, -0.0004772096872329712, -0.0009781420230865479, -0.001599244773387909, 0.0005693212151527405, 0.0001690387725830078, -0.0005792751908302307, -0.0001140497624874115 ], "p_value": 0.20205, "mde": 0.0005783116111008068, "mde_rel_pct": 1.412907813059459, "verdict": "no measurable effect", "system_worked": false }, "mechanism_signature": { "mechanism_signature": { "predicted_prior_vs_observed_load_mae": 0.12510760004321733, "predicted_prior_mean": [ 0.4308023750782013, 0.25607896596193314, 0.31311867758631706 ], "observed_load_mean": [ 0.3356913700699806, 0.3363957405090332, 0.327912874519825 ], "confirmed": false, "note": "Measured on trained MoE systems; lower MAE means the coupled prior tracked realized routing load." }, "selection": { "baseline_union_grid": [ { "lr": 0.001, "eta": 0.03 }, { "lr": 0.001, "eta": 0.1 }, { "lr": 0.001, "eta": 0.3 }, { "lr": 0.003, "eta": 0.03 }, { "lr": 0.003, "eta": 0.1 }, { "lr": 0.003, "eta": 0.3 }, { "lr": 0.006, "eta": 0.03 }, { "lr": 0.006, "eta": 0.1 }, { "lr": 0.006, "eta": 0.3 } ], "idea_grid": [ { "lr": 0.001, "eta": 0.03 }, { "lr": 0.003, "eta": 0.1 }, { "lr": 0.006, "eta": 0.3 } ], "idea_best": { "lr": 0.006, "eta": 0.3 } }, "per_seed_behavior": [ { "seed": 0, "baseline": { "metric": 0.036128461360931396, "load_var": 0.0028337526600807905, "load": [ 0.4048278331756592, 0.2771647572517395, 0.31800737977027893 ], "train_load": [ 0.4017528295516968, 0.2775239944458008, 0.32072314620018005 ], "prior": [ 0.29100728034973145, 0.3502899706363678, 0.35870274901390076 ] }, "idea": { "metric": 0.03617550805211067, "load_var": 3.5828783438773826e-05, "load": [ 0.33977699279785156, 0.3253572881221771, 0.3348657488822937 ], "train_load": [ 0.3369426131248474, 0.32563209533691406, 0.33742526173591614 ], "prior": [ 0.2513711452484131, 0.3956768214702606, 0.35295209288597107 ] } }, { "seed": 1, "baseline": { "metric": 0.04280463978648186, "load_var": 0.001193343661725521, "load": [ 0.3492494523525238, 0.3653755486011505, 0.2853749692440033 ], "train_load": [ 0.35198989510536194, 0.3660563826560974, 0.2819536626338959 ], "prior": [ 0.3014691472053528, 0.3053934872150421, 0.3931373059749603 ] }, "idea": { "metric": 0.042327430099248886, "load_var": 3.9590442611370236e-05, "load": [ 0.3245660364627838, 0.3390347957611084, 0.3363991379737854 ], "train_load": [ 0.3271709382534027, 0.339790940284729, 0.3330380916595459 ], "prior": [ 0.3157062232494354, 0.28578829765319824, 0.39850544929504395 ] } }, { "seed": 2, "baseline": { "metric": 0.04104797914624214, "load_var": 0.006669501308351755, "load": [ 0.2326611876487732, 0.43269073963165283, 0.3346480429172516 ], "train_load": [ 0.23457762598991394, 0.42982161045074463, 0.33560073375701904 ], "prior": [ 0.42684873938560486, 0.26154571771621704, 0.3116055727005005 ] }, "idea": { "metric": 0.040069837123155594, "load_var": 7.834684765839484e-06, "load": [ 0.3340980112552643, 0.3295874297618866, 0.33631452918052673 ], "train_load": [ 0.33461418747901917, 0.32943210005760193, 0.3359537124633789 ], "prior": [ 0.5657042860984802, 0.17496956884860992, 0.25932615995407104 ] } }, { "seed": 3, "baseline": { "metric": 0.047543514519929886, "load_var": 0.006999241188168526, "load": [ 0.25769320130348206, 0.29236385226249695, 0.4499429166316986 ], "train_load": [ 0.25029802322387695, 0.30129286646842957, 0.4484090209007263 ], "prior": [ 0.37379252910614014, 0.3889646530151367, 0.23724278807640076 ] }, "idea": { "metric": 0.04594426974654198, "load_var": 4.3429772631498054e-05, "load": [ 0.34203284978866577, 0.3318789601325989, 0.32608816027641296 ], "train_load": [ 0.3329542577266693, 0.344044029712677, 0.32300177216529846 ], "prior": [ 0.4113822877407074, 0.3852009177207947, 0.20341677963733673 ] } }, { "seed": 4, "baseline": { "metric": 0.038935478776693344, "load_var": 0.016052283346652985, "load": [ 0.492380827665329, 0.32526832818984985, 0.18235087394714355 ], "train_load": [ 0.49742141366004944, 0.3184661865234375, 0.18411238491535187 ], "prior": [ 0.23821482062339783, 0.34330517053604126, 0.4184800386428833 ] }, "idea": { "metric": 0.039504799991846085, "load_var": 0.00030185282230377197, "load": [ 0.33665525913238525, 0.352755606174469, 0.31058913469314575 ], "train_load": [ 0.34037309885025024, 0.34567272663116455, 0.3139541447162628 ], "prior": [ 0.1704246699810028, 0.30951765179634094, 0.520057737827301 ] } }, { "seed": 5, "baseline": { "metric": 0.04488188028335571, "load_var": 0.012242220342159271, "load": [ 0.17938412725925446, 0.4345579147338867, 0.38605797290802 ], "train_load": [ 0.18139150738716125, 0.44180044531822205, 0.3768080174922943 ], "prior": [ 0.4594753384590149, 0.2737727165222168, 0.2667520344257355 ] }, "idea": { "metric": 0.04505091905593872, "load_var": 3.8094429328339174e-05, "load": [ 0.3336072862148285, 0.3407518267631531, 0.32564085721969604 ], "train_load": [ 0.3371290862560272, 0.3452141582965851, 0.3176567256450653 ], "prior": [ 0.6375130414962769, 0.15396623313426971, 0.20852075517177582 ] } }, { "seed": 6, "baseline": { "metric": 0.04060979187488556, "load_var": 0.0106210857629776, "load": [ 0.18771888315677643, 0.4115215837955475, 0.4007594883441925 ], "train_load": [ 0.18947289884090424, 0.41077253222465515, 0.39975452423095703 ], "prior": [ 0.43537044525146484, 0.26841896772384644, 0.29621052742004395 ] }, "idea": { "metric": 0.04003051668405533, "load_var": 9.865413449006155e-05, "load": [ 0.3200598955154419, 0.33598971366882324, 0.34395039081573486 ], "train_load": [ 0.3222788870334625, 0.3344203233718872, 0.3433007597923279 ], "prior": [ 0.6139684319496155, 0.1969204843044281, 0.1891111135482788 ] } }, { "seed": 7, "baseline": { "metric": 0.03549303486943245, "load_var": 0.01686282269656658, "load": [ 0.2574109137058258, 0.5161084532737732, 0.22648058831691742 ], "train_load": [ 0.2536790072917938, 0.5173004865646362, 0.22902047634124756 ], "prior": [ 0.40807998180389404, 0.21958644688129425, 0.3723335862159729 ] }, "idea": { "metric": 0.03537898510694504, "load_var": 0.00034477459848858416, "load": [ 0.3547346293926239, 0.3358103036880493, 0.3094550371170044 ], "train_load": [ 0.3507271409034729, 0.33421364426612854, 0.31505918502807617 ], "prior": [ 0.4803489148616791, 0.14659175276756287, 0.37305933237075806 ] } } ] } }