{"id":"W4402667015","doi":"10.18653/v1/2024.acl-long.22","title":"Picturing Ambiguity: A Visual Twist on the Winograd Schema Challenge","year":2024,"lang":"en","type":"article","venue":"","topic":"Art, Technology, and Culture","field":"Arts and Humanities","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Brock University","funders":"Natural Sciences and Engineering Research Council of Canada; Canada Research Chairs","keywords":"Ambiguity; Computer science; Schema (genetic algorithms); Twist; Computer graphics (images); Artificial intelligence; Computer vision; Human–computer interaction; Information retrieval; Programming language; Mathematics; Geometry","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":[0.006378119,0.001389239,0.0009651962,0.002263748,0.001713451,0.005581389,0.002786543,0.002676701,0.007289105],"category_scores_gemma":[0.03026177,0.0006780722,0.001620576,0.002185344,0.002762561,0.011049,0.006010155,0.005144675,0.002808015],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001352338,"about_ca_system_score_gemma":0.00167433,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.009631004,"about_ca_topic_score_gemma":0.01700669,"domain_scores_codex":[0.9945516,0.002550876,0.000306066,0.001461407,0.0009393612,0.0001906488],"domain_scores_gemma":[0.9895319,0.005601446,0.0003678183,0.003339891,0.0008441072,0.0003148476],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"not_applicable","study_design_scores_codex":[0.001084853,0.0003352338,0.01969155,0.001612336,0.0003321872,0.001042093,0.005681991,0.0431165,0.008863206,0.1449068,0.2807985,0.4925347],"study_design_scores_gemma":[0.0002474588,0.0002361005,0.005406669,0.000793785,0.0001206827,0.001907873,0.005796693,0.3481776,0.02115777,0.3315686,0.2844006,0.000186123],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2383648,0.009200555,0.6076179,0.02948985,0.002228235,0.0007223623,0.03581499,0.02834617,0.04821502],"genre_scores_gemma":[0.5181724,0.001625431,0.425449,0.00442916,0.0003426159,0.0003509914,0.037932,0.003601107,0.008097315],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.009631004,"threshold_uncertainty_score":0.0337311,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04739489920513897,"score_gpt":0.2632124322978064,"score_spread":0.2158175330926675,"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."}}