{"id":"W4413149443","doi":"10.1016/j.eng.2025.07.037","title":"Can large language models solve complex engineering issues? Practical applications in reliability systems engineering","year":2025,"lang":"en","type":"article","venue":"Engineering","topic":"Software Reliability and Analysis Research","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Alberta","funders":"National Natural Science Foundation of China","keywords":"Computer science; Reliability (semiconductor); Complex system; Reliability engineering; Systems engineering; Software engineering; Engineering; 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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0009264686,0.000256858,0.0004004057,0.0005135862,0.00007260124,0.0002307489,0.000693745,0.0001352799,0.000005576929],"category_scores_gemma":[0.0006225166,0.0002733271,0.0001173271,0.001643775,0.0000132466,0.0004490854,0.0003716701,0.0005751441,0.000009526753],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003314455,"about_ca_system_score_gemma":0.00009763865,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003329433,"about_ca_topic_score_gemma":0.00001462144,"domain_scores_codex":[0.9978669,0.00003567162,0.000468632,0.0005880499,0.000375635,0.0006650535],"domain_scores_gemma":[0.9982347,0.0005184376,0.00003804725,0.0009586865,0.00009645443,0.0001535987],"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":[0.000001137144,0.0000828245,0.0003906926,0.000308769,0.00003191845,0.00001409542,0.0004099275,0.9164093,0.001151692,0.08095706,0.00006374717,0.0001788427],"study_design_scores_gemma":[0.0001934312,0.000008094627,0.001673794,0.0001011522,0.000008104448,0.000003887717,0.0000844407,0.9933871,0.000319534,0.0001084762,0.003876473,0.0002354845],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01683135,0.0007094732,0.9803028,0.0008385373,0.0001640733,0.0004367237,0.00001066515,0.0006204229,0.00008593109],"genre_scores_gemma":[0.9433993,0.00003130598,0.05608137,0.00002039586,0.00008865051,0.0002520418,0.0000173684,0.00002293703,0.00008670641],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9265679,"threshold_uncertainty_score":0.9999719,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01648185438212156,"score_gpt":0.3031423616824004,"score_spread":0.2866605073002789,"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."}}