{"id":"W4386566644","doi":"10.18653/v1/2023.eacl-main.188","title":"How Many and Which Training Points Would Need to be Removed to Flip this Prediction?","year":2023,"lang":"en","type":"article","venue":"","topic":"Machine Learning and Data Classification","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Army Research Office; Atomic Energy of Canada Limited","keywords":"Robustness (evolution); Computer science; Text categorization; Cardinality (data modeling); Categorization; Training set; Simple (philosophy); Machine learning; Artificial intelligence; Set (abstract data type); Regular polygon; Algorithm; Data mining; Mathematics","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":[],"consensus_categories":[],"category_scores_codex":[0.0005259755,0.0001018757,0.0001079199,0.0002266036,0.0001599425,0.0004287643,0.0003930379,0.00008093095,0.00002690785],"category_scores_gemma":[0.0004614693,0.00009120574,0.00001651858,0.001198496,0.000008004743,0.0003598604,0.0002677919,0.0002031555,0.0001671456],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001914423,"about_ca_system_score_gemma":0.00002839844,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00006049117,"about_ca_topic_score_gemma":0.00003918499,"domain_scores_codex":[0.998871,0.00005605923,0.0001285795,0.0004542476,0.0002491674,0.000240947],"domain_scores_gemma":[0.999113,0.00007893458,0.00003148061,0.000505032,0.00006826369,0.000203332],"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":[0.00003654152,0.00005258728,0.009038311,0.00003773454,0.00004536704,0.0000140562,0.02727228,0.0003715082,0.01990945,0.11224,0.3194075,0.5115747],"study_design_scores_gemma":[0.0004298893,0.0001858229,0.2052413,0.00003237347,0.000007203682,0.00001665706,0.001496809,0.4520524,0.000438003,0.0005202208,0.3392536,0.0003257007],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.06682457,0.000007414253,0.5344734,0.38166,0.000407659,0.0002859749,0.00001648652,0.001336766,0.01498782],"genre_scores_gemma":[0.8753262,0.000012233,0.08323118,0.007006113,0.0001627865,0.00004359088,0.00005697086,0.00001958884,0.03414129],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8085017,"threshold_uncertainty_score":0.4134584,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05024372021815765,"score_gpt":0.2772746563930931,"score_spread":0.2270309361749354,"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."}}