{ "bench_version": 1, "track": "heterogeneous_regime_moe", "model": "custom_sparse_moe", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.006, "alpha": 0.0, "aux": 0.05, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "sweep": [ { "cfg": { "lr": 0.001, "alpha": 0.0, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.3828819803893566, "per_seed": [ 0.38268065452575684, 0.42182663083076477, 0.39980241656303406, 0.4181539714336395, 0.3709557354450226, 0.3404829204082489, 0.3406737446784973, 0.3884797692298889 ] }, { "cfg": { "lr": 0.001, "alpha": 0.0, "aux": 0.02, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.38272077590227127, "per_seed": [ 0.38274234533309937, 0.42027127742767334, 0.40094804763793945, 0.41814735531806946, 0.37042883038520813, 0.34024134278297424, 0.34057343006134033, 0.38841357827186584 ] }, { "cfg": { "lr": 0.001, "alpha": 0.0, "aux": 0.05, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.38203955814242363, "per_seed": [ 0.3827840983867645, 0.41726404428482056, 0.4006299376487732, 0.4182635545730591, 0.37063243985176086, 0.3401484489440918, 0.33957916498184204, 0.387014776468277 ] }, { "cfg": { "lr": 0.001, "alpha": 0.0, "aux": 0.1, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.3819722458720207, "per_seed": [ 0.3825948238372803, 0.4162457585334778, 0.40236616134643555, 0.4189237058162689, 0.37040382623672485, 0.3401601314544678, 0.33924177289009094, 0.3858417868614197 ] }, { "cfg": { "lr": 0.003, "alpha": 0.0, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.322677806019783, "per_seed": [ 0.336835652589798, 0.3815375566482544, 0.31310850381851196, 0.36289894580841064, 0.2942163646221161, 0.27027586102485657, 0.2819417417049408, 0.34060782194137573 ] }, { "cfg": { "lr": 0.003, "alpha": 0.0, "aux": 0.02, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.3223222754895687, "per_seed": [ 0.33747565746307373, 0.37728649377822876, 0.3187519311904907, 0.3634164333343506, 0.28765812516212463, 0.26715168356895447, 0.28378787636756897, 0.3430500030517578 ] }, { "cfg": { "lr": 0.003, "alpha": 0.0, "aux": 0.05, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.3225990757346153, "per_seed": [ 0.3332405090332031, 0.37485089898109436, 0.3205052614212036, 0.36014488339424133, 0.2920096516609192, 0.2685995101928711, 0.28830453753471375, 0.34313735365867615 ] }, { "cfg": { "lr": 0.003, "alpha": 0.0, "aux": 0.1, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.32372206822037697, "per_seed": [ 0.3341885805130005, 0.374936580657959, 0.3235602080821991, 0.3638923466205597, 0.29552894830703735, 0.26977014541625977, 0.29074564576148987, 0.3371540904045105 ] }, { "cfg": { "lr": 0.006, "alpha": 0.0, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.3063250947743654, "per_seed": [ 0.340360552072525, 0.3520633578300476, 0.29556065797805786, 0.348251074552536, 0.26277485489845276, 0.275777131319046, 0.24570734798908234, 0.3301057815551758 ] }, { "cfg": { "lr": 0.006, "alpha": 0.0, "aux": 0.02, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.3056254591792822, "per_seed": [ 0.3370864689350128, 0.3551115095615387, 0.29941967129707336, 0.33801084756851196, 0.26092731952667236, 0.2828543484210968, 0.23862560093402863, 0.3329679071903229 ] }, { "cfg": { "lr": 0.006, "alpha": 0.0, "aux": 0.05, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.3030446618795395, "per_seed": [ 0.33402499556541443, 0.3469650149345398, 0.2999320328235626, 0.3352453112602234, 0.2639300525188446, 0.2693466544151306, 0.242201566696167, 0.33271166682243347 ] }, { "cfg": { "lr": 0.006, "alpha": 0.0, "aux": 0.1, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.304129621013999, "per_seed": [ 0.33261367678642273, 0.34067946672439575, 0.2965738773345947, 0.3427000343799591, 0.2676871120929718, 0.26694804430007935, 0.2444242388010025, 0.3414105176925659 ] } ], "full": { "mean": 0.3030446618795395, "std": 0.037387950188241835, "per_seed": [ 0.33402499556541443, 0.3469650149345398, 0.2999320328235626, 0.3352453112602234, 0.2639300525188446, 0.2693466544151306, 0.242201566696167, 0.33271166682243347 ], "n": 8, "signatures": [ { "load_cv": 0.562938723734806, "raw_cv": 0.16369719803202323, "pressure": [ 0.09628282487392426, 0.081390880048275, 0.08156248927116394, 0.09052031487226486, 0.108507439494133, 0.11428291350603104, 0.06951107084751129, 0.07551244646310806 ] }, { "load_cv": 0.5744562491404501, "raw_cv": 0.1282355640205095, "pressure": [ 0.07845945656299591, 0.08906298130750656, 0.10167022794485092, 0.09873484820127487, 0.07295134663581848, 0.07633225619792938, 0.11173825711011887, 0.08862103521823883 ] }, { "load_cv": 0.6500000745006104, "raw_cv": 0.18785403298311348, "pressure": [ 0.11149176955223083, 0.08400898426771164, 0.09686370939016342, 0.06456371396780014, 0.08030611276626587, 0.09944741427898407, 0.06263477355241776, 0.1182539090514183 ] }, { "load_cv": 0.617089927191566, "raw_cv": 0.15858341208470625, "pressure": [ 0.07222633063793182, 0.06700322031974792, 0.09097043424844742, 0.0720442607998848, 0.1073877140879631, 0.09829375892877579, 0.0978299006819725, 0.11181474477052689 ] }, { "load_cv": 0.4471017718279356, "raw_cv": 0.1292689442624241, "pressure": [ 0.09001918137073517, 0.10562513768672943, 0.0990309864282608, 0.06985172629356384, 0.09212449193000793, 0.09441789984703064, 0.09658448398113251, 0.06991647183895111 ] }, { "load_cv": 0.5402776598887137, "raw_cv": 0.15026783943056055, "pressure": [ 0.0705578625202179, 0.10928625613451004, 0.08838503062725067, 0.09580148756504059, 0.10527030378580093, 0.06459853053092957, 0.11027625948190689, 0.07339463382959366 ] }, { "load_cv": 0.5783597575594265, "raw_cv": 0.11876459419632313, "pressure": [ 0.0868268609046936, 0.09471505880355835, 0.11200853437185287, 0.08284948021173477, 0.08226083219051361, 0.08512511849403381, 0.09811943769454956, 0.07566508650779724 ] }, { "load_cv": 0.4230839014019498, "raw_cv": 0.09526789188308794, "pressure": [ 0.08401855081319809, 0.09309402853250504, 0.08401310443878174, 0.08053044974803925, 0.0959775522351265, 0.10520738363265991, 0.09375947713851929, 0.08096984028816223 ] } ] } }, "idea": { "mean": 0.30424891598522663, "std": 0.049616770951017984, "per_seed": [ 0.34059908986091614, 0.3489361107349396, 0.2619922161102295, 0.3619314134120941, 0.27461159229278564, 0.2645150423049927, 0.22574420273303986, 0.35566166043281555 ], "n": 8, "signatures": [ { "load_cv": 0.3511409759493393, "raw_cv": 0.17809261381483654, "pressure": [ 0.09716491401195526, 0.0805320143699646, 0.08110848069190979, 0.08782943338155746, 0.10998763889074326, 0.11640482395887375, 0.07011502236127853, 0.07442808896303177 ] }, { "load_cv": 0.5395368337588062, "raw_cv": 0.18354195356222183, "pressure": [ 0.07514716684818268, 0.08768722414970398, 0.10187951475381851, 0.10252103954553604, 0.07107587903738022, 0.07342103868722916, 0.11874773353338242, 0.08709078282117844 ] }, { "load_cv": 0.5327288508372604, "raw_cv": 0.2340563237648359, "pressure": [ 0.11677199602127075, 0.07979930937290192, 0.09878582507371902, 0.062348250299692154, 0.07951708883047104, 0.09967383742332458, 0.05981113389134407, 0.1208629459142685 ] }, { "load_cv": 0.48010414838406806, "raw_cv": 0.2041649222357629, "pressure": [ 0.07007212191820145, 0.06405238062143326, 0.09175959974527359, 0.07002781331539154, 0.11198381334543228, 0.09871535003185272, 0.09743785858154297, 0.11352145671844482 ] }, { "load_cv": 0.2900000214553521, "raw_cv": 0.14455384015921344, "pressure": [ 0.09033729881048203, 0.10671012848615646, 0.09973443299531937, 0.06994510442018509, 0.09231198579072952, 0.0953044667840004, 0.09653972834348679, 0.06668726354837418 ] }, { "load_cv": 0.3588871657819647, "raw_cv": 0.19309554993951816, "pressure": [ 0.0671059787273407, 0.11245114356279373, 0.08675403892993927, 0.09586981683969498, 0.10887758433818817, 0.06107066944241524, 0.11346158385276794, 0.07197955995798111 ] }, { "load_cv": 0.5058655738790097, "raw_cv": 0.1606991684840679, "pressure": [ 0.08636239171028137, 0.09632236510515213, 0.11678074300289154, 0.08211132884025574, 0.08124763518571854, 0.08318377286195755, 0.10013129562139511, 0.07143087685108185 ] }, { "load_cv": 0.34161379933083946, "raw_cv": 0.12798243760960235, "pressure": [ 0.08367805182933807, 0.09391549974679947, 0.08226265013217926, 0.0783558338880539, 0.09979941695928574, 0.10702546685934067, 0.09416698664426804, 0.07836645841598511 ] } ], "best_cfg": { "lr": 0.006, "alpha": 2.0, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "sweep": [ { "cfg": { "lr": 0.001, "alpha": 0.5, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.3829677514731884 }, { "cfg": { "lr": 0.001, "alpha": 1.0, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.38300157710909843 }, { "cfg": { "lr": 0.001, "alpha": 2.0, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.3818103075027466 }, { "cfg": { "lr": 0.003, "alpha": 0.5, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.3212699890136719 }, { "cfg": { "lr": 0.003, "alpha": 1.0, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.32134268060326576 }, { "cfg": { "lr": 0.003, "alpha": 2.0, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.32290053740143776 }, { "cfg": { "lr": 0.006, "alpha": 0.5, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.3090713191777468 }, { "cfg": { "lr": 0.006, "alpha": 1.0, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.3074618801474571 }, { "cfg": { "lr": 0.006, "alpha": 2.0, "aux": 0.0, "beta": 0.9, "wd": 0.0, "epochs": 3 }, "mean": 0.30424891598522663 } ] }, "comparison": { "delta_mean": 0.0012042541056871414, "idea_wins": 3, "n_pairs": 8, "per_seed_diffs": [ 0.006574094295501709, 0.0019710958003997803, -0.03793981671333313, 0.026686102151870728, 0.01068153977394104, -0.0048316121101379395, -0.016457363963127136, 0.02294999361038208 ], "p_value": 0.8799, "mde": 0.01765867919995847, "mde_rel_pct": 5.827088024067492, "verdict": "no measurable effect", "system_worked": false, "paired_delta_mean": 0.0012042541056871414, "permutation_pvalue": 0.8799 }, "mechanism_signature": { "predicted_raw_pressure_cv_reduction": 0.17827335119625737, "observed_trained_routed_load_cv": 0.42498467117208, "confirmed": false }, "custom_track": { "name": "heterogeneous_regime_moe", "file": "stage2_bench.py", "domain": "moe-routing" }, "runtime_sec": 99.08687734603882 }