{"id":"W4412945738","doi":"10.18653/v1/2025.acl-long.276","title":"Hallucination Detox: Sensitivity Dropout (SenD) for Large Language Model Training","year":2025,"lang":"en","type":"article","venue":"","topic":"Brain Tumor Detection and Classification","field":"Neuroscience","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Compute Canada; Canadian Institute for Advanced Research","keywords":"Dropout (neural networks); Computer science; Sensitivity (control systems); Training (meteorology); Language model; Artificial intelligence; Machine learning; Engineering","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":[],"consensus_categories":[],"category_scores_codex":[0.0003750375,0.00008252771,0.00009120451,0.0001060405,0.0002136213,0.0000475201,0.00006036833,0.00005376666,0.00003183948],"category_scores_gemma":[0.0007387694,0.00007985155,0.00005353284,0.0002164034,0.00002149852,0.0001356683,0.00002033982,0.00008426027,0.00001883156],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005081155,"about_ca_system_score_gemma":0.00005300769,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000005073137,"about_ca_topic_score_gemma":0.0001001144,"domain_scores_codex":[0.99918,0.00007552608,0.000135903,0.0003045468,0.0001126442,0.0001913585],"domain_scores_gemma":[0.9994618,0.0002510879,0.0000462383,0.0001704962,0.0000374118,0.00003295463],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00002128041,0.00005674098,0.00001004476,0.0000164476,0.000002079254,0.000001971245,0.0008629179,0.0002346178,0.7989764,0.1678274,0.0005208881,0.03146923],"study_design_scores_gemma":[0.0003380685,0.000009295681,0.000128213,0.000005366782,0.000005193932,0.000003608089,0.0005917072,0.7134088,0.2829793,0.001314189,0.001147926,0.00006832486],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.08857492,0.000004175302,0.8645366,0.001674817,0.0002095063,0.0002967848,0.00002248552,0.0002676775,0.04441301],"genre_scores_gemma":[0.9786863,0.000001746629,0.001910489,0.002474004,0.00002828259,0.00003776133,0.000005051123,0.000009142529,0.0168472],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8901114,"threshold_uncertainty_score":0.3256252,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05352404748894912,"score_gpt":0.3223530593030851,"score_spread":0.268829011814136,"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."}}