{"id":"W4404752296","doi":"10.18653/v1/2024.blackboxnlp-1.6","title":"LLM Internal States Reveal Hallucination Risk Faced With a Query","year":2024,"lang":"en","type":"article","venue":"","topic":"Hallucinations in medical conditions","field":"Neuroscience","cited_by":16,"is_retracted":false,"has_abstract":true,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"National Natural Science Foundation of China","keywords":"Computer science","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":["insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.0002663372,0.0001127599,0.00009031551,0.0002012966,0.0001479861,0.0001714604,0.00019381,0.00004589501,0.001259545],"category_scores_gemma":[0.0010119,0.00008007604,0.00003488032,0.0004593933,0.0001455997,0.0003624683,0.00004756267,0.0003564823,0.0005886081],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00008427746,"about_ca_system_score_gemma":0.00008285291,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00009137353,"about_ca_topic_score_gemma":0.00009234259,"domain_scores_codex":[0.9986544,0.0001438825,0.0001915504,0.000386619,0.0004228181,0.0002007665],"domain_scores_gemma":[0.9989888,0.0006438044,0.00003744101,0.0001885086,0.0000485454,0.00009289308],"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":[0.0001584937,0.001190804,0.002808012,0.000424347,0.0001678101,0.001674347,0.007601579,0.003305622,0.1286891,0.4613434,0.126074,0.2665625],"study_design_scores_gemma":[0.002363974,0.002022467,0.01199769,0.001119713,0.0002521385,0.0007531864,0.001430234,0.4519747,0.2872196,0.07703524,0.1621816,0.001649466],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5387568,0.00009205091,0.3725685,0.006842103,0.0007511344,0.0004886751,0.0001491526,0.001177512,0.07917406],"genre_scores_gemma":[0.9826339,0.00004885201,0.003282126,0.0006165852,0.00007065526,0.00004841498,0.00001212515,0.00001983746,0.01326755],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.448669,"threshold_uncertainty_score":0.9996535,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01575168994856648,"score_gpt":0.2904858905976904,"score_spread":0.2747342006491239,"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."}}