{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001809835,0.0008203514,0.000531422,0.0009964112,0.0004504156,0.001718819,0.0008188377,0.0008952146,0.001939429],"category_scores_gemma":[0.01518274,0.0003560189,0.0003037236,0.00048361,0.0007147875,0.003160626,0.00146761,0.001223633,0.000757066],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004002431,"about_ca_system_score_gemma":0.0003853045,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0008900338,"about_ca_topic_score_gemma":0.001499326,"domain_scores_codex":[0.9991546,0.0003275198,0.00007585654,0.0001895581,0.0002004092,0.0000519787],"domain_scores_gemma":[0.9923142,0.00407703,0.001241918,0.0008491764,0.001230203,0.0002874216],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.001617487,0.0001733052,0.0128783,0.001061755,0.0002050906,0.001108756,0.003150715,0.04480801,0.1704586,0.01785432,0.01242494,0.7342587],"study_design_scores_gemma":[0.0001022633,0.0008298272,0.01437377,0.0001817353,0.0002277383,0.001131299,0.00120811,0.7784382,0.1318707,0.05050707,0.0209717,0.0001576911],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1393823,0.001288534,0.8502845,0.00107866,0.0001865471,0.0001531129,0.0004622186,0.004465067,0.002699061],"genre_scores_gemma":[0.8044072,0.0004134671,0.1921497,0.0001980235,0.0001479206,0.00005337086,0.0006229961,0.0002185231,0.00178892],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.001939429,"threshold_uncertainty_score":0.009571433,"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."}}