{"id":"W3099859597","doi":"10.18653/v1/2020.coling-main.515","title":"An Analysis of Dataset Overlap on Winograd-Style Tasks","year":2020,"lang":"en","type":"preprint","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research; Microsoft Research","keywords":"Schema (genetic algorithms); Computer science; Pronoun; Artificial intelligence; Natural language processing; Style (visual arts); Parsing; Language model; Machine learning; Linguistics","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01142087,0.00113439,0.001531326,0.003224402,0.002155308,0.003286432,0.002137323,0.001537333,0.002559764],"category_scores_gemma":[0.0495263,0.0006084841,0.001380328,0.005352527,0.001866236,0.004036996,0.005685486,0.002139726,0.001703715],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00134343,"about_ca_system_score_gemma":0.00152702,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006126769,"about_ca_topic_score_gemma":0.009543615,"domain_scores_codex":[0.9808333,0.005774806,0.00198137,0.005237517,0.00478021,0.001392814],"domain_scores_gemma":[0.9628924,0.02050102,0.001840583,0.008768944,0.004528223,0.001468809],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.007973743,0.00212034,0.3077738,0.003201814,0.002722466,0.002810173,0.007524833,0.0602555,0.02710412,0.01266484,0.189224,0.3766245],"study_design_scores_gemma":[0.0009433131,0.002813128,0.4598835,0.0007790326,0.001004318,0.007445008,0.01089949,0.2244601,0.05276849,0.02329457,0.2152378,0.0004713069],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9429575,0.004829251,0.01040944,0.001139372,0.00033133,0.0001783726,0.02530575,0.002765909,0.01208299],"genre_scores_gemma":[0.8063689,0.0008169387,0.01278252,0.0004754433,0.0001442746,0.0004826342,0.1739957,0.001439696,0.003493834],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01142087,"threshold_uncertainty_score":0.06040007,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02714482607621089,"score_gpt":0.3299558031858558,"score_spread":0.3028109771096449,"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."}}