{"id":"W4417436230","doi":"10.1145/3785363","title":"Galápagos: Automated N-Version Programming with LLMs","year":2025,"lang":"en","type":"article","venue":"ACM Transactions on Software Engineering and Methodology","topic":"Software Testing and Debugging Techniques","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal","funders":"","keywords":"Correctness; Redundancy (engineering); Software; Symbolic execution; Program analysis; Automatic programming; Software quality; Programming paradigm; Software testing","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":[0.001704426,0.0007043533,0.0003685252,0.0006781533,0.0004068728,0.001046404,0.002145301,0.0007361535,0.002377114],"category_scores_gemma":[0.006068582,0.0007161617,0.001111891,0.0003301502,0.001745023,0.001925438,0.002436373,0.001170803,0.0006497688],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006688661,"about_ca_system_score_gemma":0.001059486,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0009956084,"about_ca_topic_score_gemma":0.001341541,"domain_scores_codex":[0.9983918,0.0004922008,0.0001058529,0.0003137095,0.0005590653,0.0001374139],"domain_scores_gemma":[0.9954553,0.00183395,0.0003878055,0.001922293,0.0003096346,0.000091043],"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":[0.001374636,0.0004301914,0.02462915,0.0009431271,0.0002331273,0.001577889,0.002237456,0.2815039,0.1866264,0.1045779,0.00835447,0.3875117],"study_design_scores_gemma":[0.000147734,0.0004079213,0.001708147,0.0001045774,0.00007248149,0.0007491267,0.0001248392,0.7779586,0.1464985,0.05316392,0.018973,0.00009109605],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.09482519,0.0001171833,0.8746911,0.0001343047,0.00002943262,0.0001422328,0.0001875463,0.02720347,0.002669525],"genre_scores_gemma":[0.5407792,0.00009174553,0.4516693,0.0001721053,0.00001360771,0.0002322735,0.0005250438,0.004131954,0.002384814],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.002377114,"threshold_uncertainty_score":0.009013951,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04061881690512797,"score_gpt":0.3080365130385907,"score_spread":0.2674176961334627,"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."}}