{"id":"W4206850841","doi":"10.18653/v1/2022.acl-long.236","title":"Hallucinated but Factual! Inspecting the Factuality of Hallucinations in Abstractive Summarization","year":2022,"lang":"en","type":"article","venue":"Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","topic":"Topic Modeling","field":"Computer Science","cited_by":67,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University; Mila - Quebec Artificial Intelligence Institute","funders":"Samsung; Compute Canada; Canadian Institute for Advanced Research","keywords":"Hallucinating; Automatic summarization; Computer science; Artificial intelligence; Natural language processing; Reinforcement learning; Machine learning","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":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.002447553,0.0001611776,0.000283443,0.0001371239,0.0007099733,0.00005714423,0.001416339,0.00007446804,0.000002853418],"category_scores_gemma":[0.01613513,0.0001247095,0.0002058501,0.0007739895,0.00007439483,0.0001353303,0.0006988025,0.000389062,3.123033e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004868493,"about_ca_system_score_gemma":0.0002105635,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000286001,"about_ca_topic_score_gemma":0.00003233552,"domain_scores_codex":[0.997344,0.00008896212,0.0008831274,0.0003211911,0.001107076,0.0002556],"domain_scores_gemma":[0.9932178,0.001473469,0.002442924,0.0002014867,0.002634482,0.00002980577],"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":[0.0000423665,0.0002535758,0.1851844,0.0001726295,0.000147476,1.206875e-7,0.01323694,0.6836606,0.0009341137,0.1154792,0.0002135504,0.0006750306],"study_design_scores_gemma":[0.001130654,0.0001292406,0.3873063,0.0001939836,0.000112736,0.000002691877,0.005437449,0.5532625,0.004608728,0.04666389,0.0008014902,0.0003504028],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9686272,0.00007980435,0.0129214,0.004184224,0.002257946,0.001931134,0.0005670363,0.0001157247,0.009315489],"genre_scores_gemma":[0.9945679,0.000001033054,0.004844213,0.00009257087,0.00009019362,0.00003365332,0.00001162576,0.00001537009,0.0003433976],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.2021219,"threshold_uncertainty_score":0.9921524,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01161773811128975,"score_gpt":0.237852164995372,"score_spread":0.2262344268840823,"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."}}