{"id":"W4409364220","doi":"10.1609/aaai.v39i17.33957","title":"GenPlan: Generative Sequence Models as Adaptive Planners","year":2025,"lang":"en","type":"article","venue":"Proceedings of the AAAI Conference on Artificial Intelligence","topic":"Model-Driven Software Engineering Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Generative grammar; Sequence (biology); Computer science; Artificial intelligence; Biology","routes":{"ca_aff":true,"ca_fund":true,"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":[0.0004640706,0.0003177943,0.0003252712,0.000239888,0.0001936388,0.0002130133,0.002864097,0.0001444585,0.00001476258],"category_scores_gemma":[0.00008812229,0.0002553613,0.0001265416,0.0008856553,0.0002742149,0.0006556095,0.0006116152,0.0004186569,0.0000367163],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001251048,"about_ca_system_score_gemma":0.0002414236,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00008038272,"about_ca_topic_score_gemma":0.000005466818,"domain_scores_codex":[0.9978983,0.0000270978,0.0005168273,0.0006725115,0.0004884342,0.000396858],"domain_scores_gemma":[0.9983474,0.00009697665,0.0002603101,0.0004689886,0.0007382107,0.00008810376],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00003426579,0.0000497172,0.00001419621,0.00001911728,0.00002417674,0.000001124919,0.0009412364,0.003258004,0.01311891,0.9372984,0.0003314977,0.04490938],"study_design_scores_gemma":[0.00001097715,0.0000878508,0.00000509251,0.0001892323,0.000006229093,0.00000289367,0.00005131164,0.3700992,0.296344,0.3328645,0.0001885053,0.0001502722],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.00720834,0.00003145462,0.9758953,0.001969864,0.0003334139,0.000465318,0.000009158247,0.0003712062,0.01371595],"genre_scores_gemma":[0.8724318,0.00004337161,0.1264662,0.0004202418,0.00003022881,0.00008229444,6.065409e-7,0.00001255334,0.0005127852],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.8652234,"threshold_uncertainty_score":0.9999899,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1034813939047148,"score_gpt":0.3105344415010804,"score_spread":0.2070530475963656,"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."}}