{"id":"W4413761356","doi":"10.55092/sc20250021","title":"Fine-tuning large language models and evaluating retrieval methods for improved question answering on building codes","year":2025,"lang":"en","type":"article","venue":"Smart Construction","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Alberta","funders":"","keywords":"Question answering; Computer science; Natural language processing; Information retrieval; Artificial intelligence","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":[0.004255258,0.001484937,0.001864882,0.002080278,0.0009324529,0.002281695,0.003219738,0.00248776,0.003924475],"category_scores_gemma":[0.02284961,0.0006858439,0.001652986,0.001749808,0.001135853,0.006786839,0.002153657,0.002357516,0.002474413],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001766482,"about_ca_system_score_gemma":0.002093063,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01139852,"about_ca_topic_score_gemma":0.0161895,"domain_scores_codex":[0.9961739,0.001701841,0.000251442,0.0008869309,0.0007000083,0.0002858871],"domain_scores_gemma":[0.9822858,0.01275713,0.0004310786,0.002577939,0.001630914,0.0003171301],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001700861,0.001049105,0.005350747,0.001001515,0.0003250542,0.0002101149,0.0005412966,0.2778331,0.03549864,0.01281371,0.02624368,0.6374321],"study_design_scores_gemma":[0.00006461622,0.0001041475,0.0003306694,0.00001473795,0.00005004591,0.00003926538,0.0001003448,0.9829776,0.007850518,0.007306952,0.001145885,0.00001520239],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1994211,0.001882814,0.7647059,0.0009543224,0.0002507974,0.0004225284,0.00139583,0.02746483,0.003501889],"genre_scores_gemma":[0.588392,0.0005205676,0.3996374,0.0004538108,0.0001676439,0.0003450959,0.005828702,0.001513938,0.003140738],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01139852,"threshold_uncertainty_score":0.02266437,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02152878016742676,"score_gpt":0.3888037418779764,"score_spread":0.3672749617105497,"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."}}