{"id":"W4414196974","doi":"10.1109/lcn65610.2025.11146057","title":"SIMCODE: A Benchmark for Natural Language to ns-3 Network Simulation Code Generation","year":2025,"lang":"en","type":"article","venue":"","topic":"Simulation Techniques and Applications","field":"Decision Sciences","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"Queen's University","funders":"","keywords":"Scripting language; Benchmark (surveying); Code (set theory); Focus (optics); Natural language; Natural language generation; Automation; Code generation","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":[],"consensus_categories":[],"category_scores_codex":[0.0009213267,0.00009770941,0.0001390943,0.0001728516,0.0002622213,0.0002183448,0.0003119988,0.00005570821,0.0002459158],"category_scores_gemma":[0.001121623,0.00007654005,0.00008746204,0.001040121,0.00001444235,0.0001506956,0.00006825822,0.00005352678,0.00005561952],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00004061117,"about_ca_system_score_gemma":0.00004474185,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001169964,"about_ca_topic_score_gemma":0.0001287929,"domain_scores_codex":[0.9985919,0.00004206754,0.000445071,0.0003920346,0.0003521741,0.0001767518],"domain_scores_gemma":[0.9978189,0.001177684,0.00007958659,0.0004745725,0.0003821129,0.00006714304],"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":[0.00003504422,0.00002033421,0.000691429,0.000001898094,0.000006061441,8.94405e-8,0.0001206688,0.6154654,0.001058706,0.06466006,0.2136025,0.1043377],"study_design_scores_gemma":[0.0001190508,0.00001187744,0.00226616,0.000004824949,0.000005795341,4.645851e-8,0.00005156335,0.7604941,0.0006355084,0.01950325,0.2168265,0.00008134892],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0255278,0.0001047778,0.9636441,0.003642104,0.0003742406,0.001013924,0.00002168947,0.0001532477,0.005518045],"genre_scores_gemma":[0.9351311,9.793175e-7,0.0383264,0.003318235,0.000365909,0.0001706539,0.00006194592,0.000006470442,0.02261825],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9253178,"threshold_uncertainty_score":0.3121213,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1025791354737176,"score_gpt":0.4863095816006462,"score_spread":0.3837304461269286,"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."}}