{"id":"W4417003494","doi":"10.1109/icce-asia67487.2025.11263774","title":"Action Elimination through Hallucination Detection: A Reinforcement Learning Approach for LLM Fine-Tuning","year":2025,"lang":"","type":"article","venue":"","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Nexen (Canada)","funders":"","keywords":"Reinforcement learning; Action (physics); Focus (optics); Filter (signal processing); Noise (video); Reinforcement","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"],"consensus_categories":[],"category_scores_codex":[0.001835109,0.0005513713,0.000489044,0.0006030321,0.002251295,0.0007553349,0.0008267292,0.0004398052,0.00008928998],"category_scores_gemma":[0.001416436,0.0006214312,0.0002680649,0.001888226,0.00009944792,0.002922019,0.0006055679,0.00105948,0.00001651487],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009620687,"about_ca_system_score_gemma":0.0002984288,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001853648,"about_ca_topic_score_gemma":0.00003624937,"domain_scores_codex":[0.9956675,0.0004456692,0.001036167,0.001348069,0.000732624,0.0007700236],"domain_scores_gemma":[0.9971898,0.0005055611,0.0007557356,0.0006749101,0.0007905809,0.00008340186],"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":[0.00009597292,0.00008736576,0.00002795702,0.0002832481,0.00007910969,6.450264e-7,0.001362597,0.6318107,0.000672991,0.04962307,0.00006944252,0.3158869],"study_design_scores_gemma":[0.001864026,0.0005664451,0.0002183678,0.000163201,0.0001541003,0.000008077271,0.0008618471,0.98145,0.004967436,0.001205992,0.007998005,0.0005424891],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0002776457,0.0001544511,0.953906,0.001511118,0.002626813,0.001561841,4.053384e-7,0.0004946355,0.0394671],"genre_scores_gemma":[0.7363206,0.00004855124,0.245863,0.0002025222,0.0003933141,0.0002499612,0.00004610501,0.0000324882,0.01684347],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.736043,"threshold_uncertainty_score":0.9996237,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03738149689377482,"score_gpt":0.3173500053187909,"score_spread":0.2799685084250161,"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."}}