{"id":"W3173576323","doi":"10.48550/arxiv.2106.10865","title":"Benign Overfitting in Multiclass Classification: All Roads Lead to Interpolation","year":2021,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Machine Learning and Data Classification","field":"Computer Science","cited_by":13,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Simons Institute for the Theory of Computing, University of California Berkeley; King Abdullah University of Science and Technology; National Science Foundation","keywords":"Overfitting; Support vector machine; Artificial intelligence; Computer science; Interpretability; Machine learning; Pattern recognition (psychology); Multiclass classification; Structural risk minimization; Binary classification; Multinomial logistic regression; Artificial neural network; Mathematics; Algorithm","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":[0.0005457582,0.0002825225,0.0003002498,0.0004578264,0.0001009871,0.0003429749,0.001525504,0.0002875357,0.00002001867],"category_scores_gemma":[0.0002462354,0.0003587201,0.0001201928,0.0009203469,0.00003544947,0.0006140755,0.00159063,0.0008108943,0.0001082151],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003733123,"about_ca_system_score_gemma":0.0001927553,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005064109,"about_ca_topic_score_gemma":0.0004376002,"domain_scores_codex":[0.9973259,0.0003654614,0.0003532923,0.001480817,0.0001394439,0.0003351172],"domain_scores_gemma":[0.9975922,0.000159314,0.0003084163,0.001602577,0.000179603,0.0001578693],"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":[0.0001063034,0.0006390256,0.1430487,0.00024323,0.0001622572,0.0004188179,0.007293143,0.4991113,0.005910655,0.3018295,0.001101818,0.0401352],"study_design_scores_gemma":[0.0003001914,0.00002629826,0.06033358,0.0001361155,0.00001712434,0.000002387923,0.0003584225,0.9360669,0.00007136891,0.0009098983,0.001419351,0.0003583632],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2803424,0.00002893815,0.715486,0.001013386,0.000475555,0.0002519967,0.000005854873,0.0002007207,0.002195033],"genre_scores_gemma":[0.9857992,0.00005199512,0.0128856,0.0002493832,0.00007425247,0.000003347019,0.0001655445,0.0000175866,0.0007530838],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.7054567,"threshold_uncertainty_score":0.9998865,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1198908800877382,"score_gpt":0.2356930585911815,"score_spread":0.1158021785034433,"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."}}