{"id":"W7106480312","doi":"10.1609/aaaiss.v7i1.36923","title":"Fine-Tuning Large Language Models for Structured ClinicalReport Generation Using GRPO","year":2025,"lang":"","type":"article","venue":"Proceedings of the AAAI Symposium Series","topic":"Topic Modeling","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"International Institute for Sustainable Development; Western University","funders":"","keywords":"Relevance (law); Adaptation (eye); Language model; Disk formatting; Medical care; Baseline (sea); Complement (music); English language","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.004664051,0.001661649,0.0009829941,0.0008324966,0.0003392246,0.001417259,0.002314899,0.001692158,0.003183582],"category_scores_gemma":[0.01903844,0.0006671295,0.001097672,0.0007020774,0.000692154,0.002235583,0.001500008,0.003263293,0.002366609],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001611494,"about_ca_system_score_gemma":0.003240456,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008011099,"about_ca_topic_score_gemma":0.01360967,"domain_scores_codex":[0.9981039,0.0008628741,0.0001209071,0.0005358977,0.000241599,0.0001348105],"domain_scores_gemma":[0.9924067,0.00536517,0.0003448718,0.0008912735,0.0006969343,0.0002951588],"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":[0.000722271,0.0006472972,0.005863404,0.000432926,0.0002138154,0.0003504745,0.0002840988,0.6532741,0.008091236,0.004141379,0.0292421,0.2967369],"study_design_scores_gemma":[0.00008569039,0.00007537634,0.0001970581,0.00001686483,0.00002192392,0.00004302407,0.00003248918,0.9909603,0.003516591,0.003191792,0.001843203,0.00001575673],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1469551,0.003007855,0.775817,0.004099662,0.0007984856,0.0007548276,0.004405514,0.05724932,0.006912206],"genre_scores_gemma":[0.6836118,0.0005925514,0.2998169,0.001986136,0.0001976695,0.0006623388,0.0074713,0.001254532,0.004406798],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.008011099,"threshold_uncertainty_score":0.02466613,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03647737372343484,"score_gpt":0.2966372792713703,"score_spread":0.2601599055479355,"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."}}