{"id":"W7133211079","doi":"10.1109/ocit66168.2025.11399871","title":"The Calibration Gap: Model-Specific Confidence Thresholds for Reliable Customer Service LLMs","year":2025,"lang":"","type":"article","venue":"","topic":"Software System Performance and Reliability","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Marriott International (Canada)","funders":"","keywords":"Reliability (semiconductor); Calibration; Service (business); Benchmark (surveying); Quality (philosophy); Automation; Service quality; Customer satisfaction","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":["metaepi_narrow","sts","scholarly_communication"],"consensus_categories":[],"category_scores_codex":[0.002285128,0.0004714458,0.0004837062,0.0001405945,0.002128233,0.001422173,0.002465534,0.0004082624,0.00005921166],"category_scores_gemma":[0.00007916441,0.0003129226,0.0002641867,0.001805505,0.0002608338,0.001704026,0.0006055293,0.0003806551,0.0002445997],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002149739,"about_ca_system_score_gemma":0.001141554,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002571182,"about_ca_topic_score_gemma":0.0001908107,"domain_scores_codex":[0.9957896,0.0001139169,0.001226937,0.001227987,0.0006643605,0.0009771247],"domain_scores_gemma":[0.995,0.0007769552,0.0002898804,0.002498145,0.001270032,0.0001649856],"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":[0.0002603259,0.0002056237,0.001912503,0.0009210624,0.000110878,0.000001365697,0.001761776,0.09271402,0.0003660186,0.6659516,0.2149554,0.0208394],"study_design_scores_gemma":[0.0006315409,0.00006104273,0.0001496646,0.0002364728,0.0000283777,0.000004216425,0.0002027746,0.8974531,0.001902694,0.02122627,0.07773484,0.0003689506],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.003102277,0.003027825,0.956033,0.01609227,0.005037617,0.002191307,0.00001311305,0.0003065487,0.01419606],"genre_scores_gemma":[0.9447721,0.002056049,0.01107521,0.004462215,0.0002853349,0.0005537924,0.000009485334,0.00003256342,0.03675331],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9449578,"threshold_uncertainty_score":0.9999323,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03205379937130462,"score_gpt":0.2816373459233823,"score_spread":0.2495835465520777,"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."}}