{"id":"W3213378669","doi":"10.1109/tit.2023.3320098","title":"Benign Overfitting in Multiclass Classification: All Roads Lead to Interpolation","year":2023,"lang":"en","type":"article","venue":"IEEE Transactions on Information Theory","topic":"Machine Learning and Data Classification","field":"Computer Science","cited_by":6,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"Natural Sciences and Engineering Research Council of Canada; King Abdullah University of Science and Technology; Google; National Science Foundation","keywords":"Overfitting; Interpolation (computer graphics); Artificial intelligence; Pattern recognition (psychology); Computer science; Lead (geology); Machine learning; Mathematics; Algorithm; Data mining; Artificial neural network; Geology; Image (mathematics)","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.001261724,0.000150699,0.0001247239,0.0009660172,0.0001860177,0.0002285494,0.0004854519,0.00009794418,0.00003122343],"category_scores_gemma":[0.00008615854,0.0001558309,0.00005824207,0.00134791,0.00002366012,0.002423202,0.000006946678,0.0003096096,0.002461078],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001299465,"about_ca_system_score_gemma":0.00004887625,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002823317,"about_ca_topic_score_gemma":0.00002042395,"domain_scores_codex":[0.9984317,0.0002043729,0.0005300802,0.0002498348,0.0003219235,0.0002620797],"domain_scores_gemma":[0.9987614,0.0003219682,0.0001525555,0.0005837301,0.00009027549,0.00009004129],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000100331,0.00009374992,0.00014652,0.00003358631,0.00002239262,0.000001238622,0.01255325,0.08166648,0.002353508,0.09114589,0.001121123,0.8107619],"study_design_scores_gemma":[0.0004871374,0.00008114731,0.00889129,0.0000465736,0.000005544223,0.000003830241,0.001017964,0.9757869,0.001643235,0.001592626,0.01020483,0.0002389274],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.008948492,0.000002021618,0.9836923,0.002387334,0.0006401631,0.0002942334,0.00001457895,0.0006126398,0.00340819],"genre_scores_gemma":[0.9922401,0.00001364072,0.00617737,0.0009553811,0.00002496427,0.0001209548,0.00005697785,0.000009392307,0.0004012201],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9832916,"threshold_uncertainty_score":0.9983156,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02861077349041017,"score_gpt":0.2886332035168037,"score_spread":0.2600224300263935,"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."}}