{"id":"W4414968374","doi":"10.48550/arxiv.2509.03378","title":"Understanding and Improving Shampoo and SOAP via Kullback-Leibler Minimization","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Speech and dialogue systems","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"RIKEN; University of Central Florida; Government of Canada; Canadian Institute for Advanced Research","keywords":"Shampoo; Overhead (engineering); Minification; Divergence (linguistics); SOAP","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"],"consensus_categories":[],"category_scores_codex":[0.0003251163,0.0002752724,0.0003314157,0.0002355324,0.0001948441,0.0003560269,0.0004505551,0.0002915301,0.000005870026],"category_scores_gemma":[0.00008588221,0.0002747015,0.00005740515,0.0002243135,0.00006869883,0.0002882603,0.001535769,0.0003101104,0.00001221559],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001338276,"about_ca_system_score_gemma":0.0001704819,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002542436,"about_ca_topic_score_gemma":0.00007564524,"domain_scores_codex":[0.9982349,0.00008488013,0.0003327006,0.0008340363,0.0001989776,0.0003144622],"domain_scores_gemma":[0.9988527,0.0001376047,0.0002032342,0.0006056201,0.00006551811,0.0001353148],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00007432143,0.0001451843,0.9207353,0.003244656,0.0003839645,0.0001062384,0.007210302,0.0004857933,0.005514547,0.03019219,0.005147468,0.02676005],"study_design_scores_gemma":[0.005948422,0.0004959388,0.286867,0.004252139,0.0003848432,0.0001876625,0.001604614,0.5875087,0.006825888,0.09463429,0.006003589,0.005286854],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1396369,0.001180096,0.8524476,0.0006487664,0.002085727,0.0004969292,0.00001432812,0.0002531338,0.003236419],"genre_scores_gemma":[0.9898231,0.0001591995,0.008839046,0.0002610537,0.0002172726,0.00002586766,0.00002719437,0.00001488648,0.0006323709],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8501862,"threshold_uncertainty_score":0.9999705,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1053909708859754,"score_gpt":0.2626024331985078,"score_spread":0.1572114623125324,"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."}}