{"id":"W4406733441","doi":"10.48550/arxiv.2501.10858","title":"Reliable Text-to-SQL with Adaptive Abstention","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Distributed and Parallel Computing Systems","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Computer science; SQL; Database","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0003917507,0.0003833529,0.0004636074,0.0002008037,0.0001775154,0.0002697998,0.001831403,0.0002464831,0.00000990304],"category_scores_gemma":[0.00003201894,0.0003365688,0.0001480531,0.0006078245,0.00003291884,0.0001556028,0.001731886,0.0005839281,0.0005455638],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001397743,"about_ca_system_score_gemma":0.0003614775,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003105953,"about_ca_topic_score_gemma":0.00002541067,"domain_scores_codex":[0.9974915,0.00009827273,0.0004261349,0.001120508,0.0003948183,0.0004687702],"domain_scores_gemma":[0.9977148,0.00007979543,0.0002775794,0.001458406,0.0003055736,0.0001638022],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"observational","study_design_scores_codex":[0.0002736013,0.0009207511,0.3462756,0.001727678,0.001476182,0.0004498421,0.007306074,0.4714675,0.0003416063,0.05532439,0.09590495,0.01853183],"study_design_scores_gemma":[0.003240032,0.001730666,0.5134323,0.01387715,0.0003277884,0.0001072102,0.0005339931,0.2163953,0.001091671,0.007136882,0.2366862,0.005440851],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1256687,0.0002263509,0.8456917,0.0009179938,0.00253666,0.0007845921,0.00007501099,0.0006764576,0.02342252],"genre_scores_gemma":[0.9759555,0.000009236305,0.01185187,0.0003894095,0.0002220034,0.00008367516,0.00006824762,0.00001671877,0.01140333],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8502868,"threshold_uncertainty_score":0.9999086,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0346877412721206,"score_gpt":0.2570595838781645,"score_spread":0.2223718426060439,"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."}}