{"id":"W4405064339","doi":"10.1115/1.4067333","title":"DesignQA: A Multimodal Benchmark for Evaluating Large Language Models’ Understanding of Engineering Documentation","year":2024,"lang":"en","type":"article","venue":"Journal of Computing and Information Science in Engineering","topic":"Design Education and Practice","field":"Engineering","cited_by":29,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"Directorate for Engineering","keywords":"Documentation; Benchmark (surveying); Computer science; Software engineering; Natural language processing; Engineering drawing; Systems engineering; Artificial intelligence; Engineering; Programming 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":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.009429027,0.002260026,0.0008222288,0.003326779,0.0006873652,0.003608623,0.003405015,0.002574675,0.01080873],"category_scores_gemma":[0.06131488,0.000541123,0.001249854,0.001598359,0.001028693,0.00376113,0.004589919,0.002053688,0.005219237],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001672424,"about_ca_system_score_gemma":0.002063085,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005628231,"about_ca_topic_score_gemma":0.004858358,"domain_scores_codex":[0.9857157,0.007708244,0.001365132,0.001507394,0.003340516,0.0003630625],"domain_scores_gemma":[0.9616728,0.02535686,0.001687134,0.004706624,0.005420343,0.001156248],"domain_codex":null,"domain_gemma":"evaluation","domain_candidate":"evaluation","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.002611439,0.003797078,0.02017171,0.005179011,0.0007046108,0.0006838968,0.003816077,0.1727208,0.03141261,0.01679222,0.155544,0.5865666],"study_design_scores_gemma":[0.0007748829,0.002642497,0.01211444,0.0007576947,0.0001868013,0.0004898169,0.00199083,0.8161103,0.03956253,0.02269893,0.1023623,0.0003090753],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3895189,0.004012353,0.4098367,0.002013891,0.001013204,0.002541753,0.02930272,0.1064928,0.05526764],"genre_scores_gemma":[0.6329985,0.0006293804,0.2753242,0.0008668064,0.0001175661,0.002296676,0.07453239,0.004947885,0.00828654],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.990571,"threshold_uncertainty_score":0.04986602,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03288544794624475,"score_gpt":0.3343424033057408,"score_spread":0.301456955359496,"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."}}