{"id":"W4396955972","doi":"10.1017/pds.2024.204","title":"Integrating large language models for improved failure mode and effects analysis (FMEA): a framework and case study","year":2024,"lang":"en","type":"article","venue":"Proceedings of the Design Society","topic":"Software Engineering Research","field":"Computer Science","cited_by":20,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université du Québec à Rimouski","funders":"","keywords":"Failure mode and effects analysis; Computer science; Reliability engineering; Natural language processing; Engineering","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":[0.007061291,0.00105416,0.0005305884,0.001694224,0.0008999094,0.001755498,0.001854303,0.001534079,0.001511154],"category_scores_gemma":[0.01191896,0.0005802562,0.000982973,0.00108977,0.001598143,0.00214019,0.001806624,0.001430867,0.0002643679],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001708446,"about_ca_system_score_gemma":0.001740744,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01030905,"about_ca_topic_score_gemma":0.01260566,"domain_scores_codex":[0.9954104,0.003044649,0.0001688409,0.0002168882,0.0009715502,0.0001876542],"domain_scores_gemma":[0.9858317,0.0113155,0.0005657006,0.001149595,0.000982162,0.0001553339],"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":[0.000820265,0.00327309,0.01208492,0.0009937462,0.0001652739,0.003276977,0.006871545,0.523025,0.05616103,0.09039477,0.004296062,0.2986373],"study_design_scores_gemma":[0.0001211162,0.0005837962,0.001981377,0.0001889013,0.0001020885,0.0004841928,0.000699417,0.945348,0.02441046,0.01185093,0.01413328,0.00009641403],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1817595,0.0002561416,0.8078155,0.0005815474,0.00002129885,0.0006839015,0.0001767769,0.003193709,0.00551167],"genre_scores_gemma":[0.3701337,0.0001333136,0.6278116,0.00005863622,0.0000101534,0.0002708162,0.000145646,0.0002536601,0.001182421],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01030905,"threshold_uncertainty_score":0.0373441,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01332308256745129,"score_gpt":0.290211175171413,"score_spread":0.2768880926039617,"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."}}