{"id":"W7119557480","doi":"","title":"Benchmarking and Adapting On-Device LLMs for Clinical Decision Support","year":2025,"lang":"","type":"article","venue":"ArXiv.org","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Alliance de recherche numérique du Canada; Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Benchmarking; Adaptability; Decision support system; Inference; Clinical decision support system; Software deployment; Benchmark (surveying); Grading (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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.006907946,0.001685359,0.0009742077,0.0014331,0.0004367518,0.002732714,0.00335518,0.002564714,0.006371079],"category_scores_gemma":[0.04700464,0.0007205368,0.001439765,0.0009654048,0.0008941874,0.002396514,0.003149751,0.002584068,0.002809083],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002380931,"about_ca_system_score_gemma":0.003655007,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01339247,"about_ca_topic_score_gemma":0.01696721,"domain_scores_codex":[0.9951684,0.002443218,0.0003956311,0.001014096,0.000737533,0.0002411285],"domain_scores_gemma":[0.98213,0.01295213,0.0005339938,0.002502268,0.00116198,0.0007195903],"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":[0.002620671,0.001072805,0.01960976,0.001189283,0.001142053,0.0007313746,0.0004903376,0.5814255,0.003849065,0.006592729,0.05611458,0.3251618],"study_design_scores_gemma":[0.0003325197,0.0003297629,0.001559515,0.0001276116,0.00009084953,0.0002021875,0.0001355872,0.9706966,0.003941216,0.0125297,0.009991824,0.00006252529],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3314554,0.007985946,0.5256073,0.01174854,0.002201058,0.001632951,0.01575352,0.08058759,0.02302778],"genre_scores_gemma":[0.7969854,0.001417934,0.1782358,0.003945796,0.0002740299,0.0005456351,0.01304451,0.001956549,0.003594354],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01339247,"threshold_uncertainty_score":0.03653312,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.4255715222243194,"score_gpt":0.5466994545321749,"score_spread":0.1211279323078555,"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."}}