{"id":"W4416613133","doi":"10.48550/arxiv.2505.23725","title":"MuLoCo: Muon is a practical inner optimizer for DiLoCo","year":2025,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Machine Learning in Materials Science","field":"Materials Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Fonds de recherche du Québec – Nature et technologies; Canada Excellence Research Chairs, Government of Canada; Compute Canada; Canadian Institute for Advanced Research","keywords":"Hyperparameter; Muon; Factor (programming language); Scale (ratio); Hyperparameter optimization; Language model; Synchronization (alternating current)","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.001155532,0.0004803363,0.0006116367,0.0002658922,0.0003360725,0.0003105843,0.001397222,0.0004955774,0.002059066],"category_scores_gemma":[0.0007313055,0.00050181,0.0002705334,0.000375983,0.0003556452,0.000424743,0.002133545,0.0005906044,0.0004021533],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002477471,"about_ca_system_score_gemma":0.0005359704,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002770745,"about_ca_topic_score_gemma":0.00001843657,"domain_scores_codex":[0.9966173,0.0003458478,0.0003898994,0.001780861,0.0001880709,0.0006779757],"domain_scores_gemma":[0.9971737,0.0005338144,0.0004646869,0.001285244,0.0003327147,0.0002098224],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001044697,0.0004791154,0.001440497,0.001550685,0.00009898133,0.0002449114,0.0007601388,0.868562,0.02455224,0.0781208,0.02292624,0.0002196451],"study_design_scores_gemma":[0.002221386,0.0002430383,0.0003195457,0.0005501346,0.0005120991,0.00001513786,0.0001745765,0.8709159,0.02719929,0.04205929,0.0539289,0.001860701],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4375588,0.00006513979,0.5453848,0.001446173,0.004104711,0.002034186,0.0005749368,0.0005865908,0.008244606],"genre_scores_gemma":[0.9237969,0.00009530341,0.05610587,0.001024222,0.0003249742,0.00002942588,0.00005223302,0.00005121075,0.01851989],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.489279,"threshold_uncertainty_score":0.9997433,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08419056832393303,"score_gpt":0.2543852188979551,"score_spread":0.1701946505740221,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}