{"id":"W4412889548","doi":"10.18653/v1/2025.acl-short.15","title":"Improving the Calibration of Confidence Scores in Text Generation Using the Output Distribution’s Characteristics","year":2025,"lang":"en","type":"article","venue":"","topic":"Advanced Text Analysis Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Institut de Valorisation des Données; Nvidia","keywords":"Calibration; Computer science; Confidence interval; Statistics; Distribution (mathematics); Mathematics","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.02105732,0.002466307,0.001650694,0.005276862,0.0009738207,0.004944759,0.002324783,0.003844985,0.002421028],"category_scores_gemma":[0.1706084,0.0007416751,0.001299349,0.003194289,0.001506707,0.00628411,0.003078241,0.004440263,0.002226517],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001859006,"about_ca_system_score_gemma":0.001595174,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003634945,"about_ca_topic_score_gemma":0.003437508,"domain_scores_codex":[0.984328,0.00694658,0.001450464,0.003807414,0.002789328,0.0006782302],"domain_scores_gemma":[0.8926799,0.07329357,0.005910763,0.01465557,0.01196828,0.00149191],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.002231584,0.0006972505,0.07648181,0.001375443,0.0008199138,0.0004044512,0.001423785,0.3142522,0.01691163,0.008981073,0.03123804,0.5451829],"study_design_scores_gemma":[0.0001572631,0.0003060292,0.01225039,0.0001546122,0.0001119995,0.0002743571,0.0002001768,0.9469085,0.01605692,0.01834208,0.005096349,0.0001412658],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.226998,0.005127453,0.7253336,0.001908659,0.0005629196,0.0005809448,0.005757852,0.02603888,0.007691664],"genre_scores_gemma":[0.8536234,0.0003623253,0.1328592,0.0005256571,0.0001851373,0.0003674609,0.00941267,0.001400155,0.001264031],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.02105732,"threshold_uncertainty_score":0.1113631,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02464011149066286,"score_gpt":0.2866724729756874,"score_spread":0.2620323614850246,"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."}}