{"id":"W7137121728","doi":"10.1109/phm-xian66756.2025.11427646","title":"How Much Confidence Do We Have for GenAI-Based Reasoningƒ A Statistical Inference Perspective for Reliability Estimation","year":2025,"lang":"","type":"article","venue":"","topic":"Software Reliability and Analysis Research","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Future Earth","funders":"","keywords":"Reliability (semiconductor); Perspective (graphical); Fiducial inference; Statistical inference; Estimation; Inference; Interval estimation; Point estimation","routes":{"ca_aff":true,"ca_fund":false,"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","metaepi_narrow","scholarly_communication"],"consensus_categories":[],"category_scores_codex":[0.003560432,0.0005882162,0.0009812426,0.0004931531,0.001088845,0.0026849,0.001723794,0.000424171,0.0001254293],"category_scores_gemma":[0.05176446,0.0005489454,0.0005864285,0.001478226,0.001135877,0.0009579904,0.0004410957,0.0005792801,0.00001475954],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001194371,"about_ca_system_score_gemma":0.004365326,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005551333,"about_ca_topic_score_gemma":0.0001845838,"domain_scores_codex":[0.9935097,0.0005998923,0.0009430514,0.002602835,0.001099777,0.001244773],"domain_scores_gemma":[0.9696109,0.02148752,0.0003097618,0.001986675,0.006183003,0.0004221206],"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":[0.00043993,0.0006392992,0.00114435,0.001519679,0.0001557964,0.000002992269,0.000787976,0.002382819,0.00005529845,0.8615919,0.001785537,0.1294944],"study_design_scores_gemma":[0.001113077,0.0005609014,0.0004141733,0.0004054984,0.0001254789,7.309835e-7,0.001072833,0.6666111,0.001934221,0.3262988,0.001089063,0.0003740794],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0005017958,0.0009231784,0.9435816,0.0503655,0.0003089694,0.00351149,0.000285362,0.000183874,0.0003382396],"genre_scores_gemma":[0.6362447,0.0001155594,0.360893,0.000194174,0.00005209039,0.0008000507,0.00002314898,0.00001504145,0.001662188],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.6642283,"threshold_uncertainty_score":0.9996962,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03371601620258263,"score_gpt":0.3781302341616838,"score_spread":0.3444142179591012,"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."}}