{"id":"W4385015322","doi":"10.48550/arxiv.2307.10236","title":"Look Before You Leap: An Exploratory Study of Uncertainty Measurement for Large Language Models","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Software Engineering Research","field":"Computer Science","cited_by":17,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"JST-Mirai Program; Japan Society for the Promotion of Science; Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Trustworthiness; Estimation; Computer science; Computer security; Economics","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.001437829,0.0003130058,0.0004143436,0.0004804571,0.0001160028,0.00007143943,0.002500522,0.0001944665,0.00000438207],"category_scores_gemma":[0.0002040615,0.0003572404,0.0001707035,0.0006354139,0.00003711212,0.0004498172,0.002279531,0.0004644028,0.00001520645],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004506657,"about_ca_system_score_gemma":0.0003516528,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000259193,"about_ca_topic_score_gemma":0.001217283,"domain_scores_codex":[0.9974307,0.0001668468,0.000254197,0.001211167,0.0004166383,0.0005204678],"domain_scores_gemma":[0.9968693,0.0001646564,0.0001599301,0.002014418,0.0005876251,0.0002040316],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00004137067,0.0006512111,0.002946673,0.0002436176,0.0002104207,0.0001798478,0.01135811,0.9728154,0.00003144314,0.01107587,0.000180024,0.000266032],"study_design_scores_gemma":[0.001352887,0.000569776,0.001268516,0.0001099653,0.00006025574,4.829583e-7,0.005372881,0.981117,0.00007919621,0.00962464,0.00002994379,0.0004144452],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5123723,0.00003069046,0.486064,0.00001083662,0.0002884787,0.0007663319,0.00004068244,0.0004137992,0.00001284792],"genre_scores_gemma":[0.9987903,0.00000975976,0.0006407751,0.000009041422,0.000063535,0.0000195618,0.00002273441,0.00004714975,0.0003972073],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4864179,"threshold_uncertainty_score":0.9998879,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1721652541498987,"score_gpt":0.244593140594098,"score_spread":0.07242788644419926,"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."}}