{"id":"W4414004064","doi":"10.14778/3749646.3749702","title":"ThriftLLM: On Cost-Effective Selection of Large Language Models for Classification Queries","year":2025,"lang":"en","type":"article","venue":"Proceedings of the VLDB Endowment","topic":"Data Quality and Management","field":"Decision Sciences","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"Selection (genetic algorithm); Computer science; Natural language processing; Artificial intelligence; Information retrieval","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":[0.002321837,0.0001006691,0.0002161458,0.0001905689,0.0001298107,0.00006280022,0.0005679149,0.00004311313,0.0000158366],"category_scores_gemma":[0.001089548,0.00006373677,0.0001210684,0.0005593011,0.00006504696,0.0003005759,0.0002024984,0.00006840115,0.000003660653],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00007445316,"about_ca_system_score_gemma":0.00002401068,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00003784867,"about_ca_topic_score_gemma":0.00002797546,"domain_scores_codex":[0.9985061,0.00002650302,0.0004304464,0.0002969868,0.0005843909,0.0001555963],"domain_scores_gemma":[0.9985948,0.0003671311,0.0004082665,0.0001882304,0.0004211348,0.00002045828],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0002435298,0.0002729058,0.0005469158,0.0001265406,0.00006211047,1.325265e-8,0.001672844,0.00009448276,0.02249976,0.9178571,0.0211655,0.03545831],"study_design_scores_gemma":[0.001709146,0.0004166032,0.008408657,0.0002607367,0.0001141703,4.042118e-7,0.01717654,0.01797265,0.6321599,0.2508278,0.07076044,0.0001928704],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5803645,0.0003804481,0.1555294,0.01975534,0.002064,0.02533662,0.0009305194,0.0002262028,0.2154129],"genre_scores_gemma":[0.9961619,0.00001513299,0.0006128923,0.000320232,0.00001958078,0.0006632246,0.000004491608,0.00000491643,0.002197652],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6670293,"threshold_uncertainty_score":0.259911,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09834815259862227,"score_gpt":0.4145340017749151,"score_spread":0.3161858491762928,"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."}}