{"id":"W4405259287","doi":"10.1007/s10664-024-10597-8","title":"Harnessing pre-trained generalist agents for software engineering tasks","year":2024,"lang":"en","type":"article","venue":"Empirical Software Engineering","topic":"Software Engineering Research","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"Polytechnique Montréal","funders":"","keywords":"Computer science; Artificial intelligence; Reinforcement learning; Software engineering; Generalizability theory; Scheduling (production processes); Machine learning; Software; Domain (mathematical analysis); Human–computer interaction; Engineering; Operations management","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":[0.001300875,0.0006634757,0.0004457176,0.0004540628,0.0004292776,0.001167753,0.001231133,0.001068769,0.006069298],"category_scores_gemma":[0.007751285,0.0004621624,0.0002924516,0.0003831373,0.0004805688,0.001301208,0.002097853,0.001427336,0.002894831],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004706619,"about_ca_system_score_gemma":0.001142163,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002151345,"about_ca_topic_score_gemma":0.005902473,"domain_scores_codex":[0.9993656,0.0001990951,0.00003048764,0.0001933952,0.0001358519,0.00007560098],"domain_scores_gemma":[0.9962292,0.001499907,0.0002870033,0.0009911491,0.0006378599,0.0003547734],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.001141315,0.002910708,0.03333971,0.0005447266,0.0002435325,0.000576374,0.003217941,0.07639164,0.1258015,0.006680654,0.009739084,0.7394127],"study_design_scores_gemma":[0.0003106714,0.002478289,0.02958281,0.0001689736,0.0003107516,0.0006281165,0.002090572,0.8290638,0.06173095,0.02078808,0.05270581,0.0001412225],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6280296,0.0003782523,0.3276268,0.0007724147,0.0001902573,0.0005103123,0.0001853449,0.005717003,0.03658989],"genre_scores_gemma":[0.8741548,0.0001976077,0.1099425,0.0004115148,0.00004997045,0.0001879581,0.0003745756,0.0002605518,0.01442053],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.006069298,"threshold_uncertainty_score":0.02030385,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03556022016790315,"score_gpt":0.3148634202315637,"score_spread":0.2793032000636606,"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."}}