{"id":"W4402683840","doi":"10.18653/v1/2024.findings-acl.19","title":"Are self-explanations from Large Language Models faithful?","year":2024,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":21,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Alliance de recherche numérique du Canada; Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Computer science; Natural language processing; Language model; Artificial intelligence; Linguistics; Philosophy","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":[0.00009406807,0.0000758177,0.00007373872,0.00007346262,0.00005301849,0.0002483907,0.0004312023,0.00004020667,0.0001141848],"category_scores_gemma":[0.00000745276,0.00006360148,0.00003997139,0.0001787884,0.000002996475,0.0005742635,0.0001644362,0.00009650359,0.0002593362],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002963899,"about_ca_system_score_gemma":0.00003066659,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001417683,"about_ca_topic_score_gemma":0.0001232656,"domain_scores_codex":[0.9992124,0.0000183202,0.0001124668,0.0003138764,0.00017352,0.0001693915],"domain_scores_gemma":[0.9994433,0.00005832229,0.00001883768,0.0004057266,0.00002124014,0.00005258355],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[2.819392e-7,0.00004596292,0.0001686298,0.00001646835,0.00004287111,0.0001344777,0.007135443,0.002223247,0.0001166535,0.9747837,0.008288468,0.007043803],"study_design_scores_gemma":[0.00005917456,0.000002241829,0.00009874262,0.00001591376,0.000003957719,0.000002272577,0.0002348844,0.9759264,0.0001474576,0.01900822,0.004411478,0.00008923755],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01501158,0.0005888825,0.9670757,0.001553849,0.0003377839,0.00005693767,0.00002097085,0.001185921,0.01416836],"genre_scores_gemma":[0.8861955,0.000005996707,0.1117318,0.0006204539,0.0001148642,0.000008726574,0.000005972128,0.000007046274,0.001309575],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9737031,"threshold_uncertainty_score":0.3333328,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02192689832023895,"score_gpt":0.2564935360003454,"score_spread":0.2345666376801065,"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."}}