{"id":"W4417077291","doi":"10.48550/arxiv.2505.23027","title":"Diverse Prototypical Ensembles Improve Robustness to Subpopulation Shift","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Domain Adaptation and Few-Shot Learning","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Canadian Institutes of Health Research; Natural Sciences and Engineering Research Council of Canada; Government of Canada; Canadian Institute for Advanced Research","keywords":"Robustness (evolution); Training set; Minification; Ensemble learning; Feature (linguistics); Pattern recognition (psychology); Extractor; Empirical risk minimization; Support vector machine","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":[0.0003799193,0.0003210201,0.0003425191,0.0002580804,0.0002192042,0.0002687881,0.001233962,0.0002998384,0.00004212285],"category_scores_gemma":[0.0002183713,0.0003283258,0.0001578007,0.0003872327,0.00003687356,0.000276415,0.001993751,0.0006709256,0.0001959978],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001293602,"about_ca_system_score_gemma":0.0002375183,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001552316,"about_ca_topic_score_gemma":0.00008943141,"domain_scores_codex":[0.9976488,0.000171319,0.0004194663,0.0009832609,0.0003863075,0.0003908004],"domain_scores_gemma":[0.998355,0.0001005431,0.0002031091,0.0009841101,0.0001597781,0.0001974191],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0002762865,0.00047755,0.5288526,0.0009021295,0.0002418261,0.0001403264,0.01040739,0.2152172,0.001430375,0.05354233,0.001945429,0.1865666],"study_design_scores_gemma":[0.0009639638,0.0001549251,0.8035396,0.0004937192,0.00005658893,0.000003439317,0.0002451397,0.1760927,0.001487706,0.002766607,0.01291171,0.001283799],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4163596,0.00002821423,0.5783371,0.001434903,0.001394355,0.001079194,0.000006724958,0.0003303004,0.001029645],"genre_scores_gemma":[0.9705441,0.000006583666,0.02620471,0.0006786759,0.0002120645,0.0002988514,0.00003343206,0.00001667815,0.002004889],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.5541845,"threshold_uncertainty_score":0.9999169,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05673509766038131,"score_gpt":0.3015717884706264,"score_spread":0.2448366908102451,"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."}}