{"id":"W7117116563","doi":"10.2139/ssrn.5960744","title":"Retrieval at Risk: Systematic Failures in Embedding-Based Retrieval for Nuclear Regulatory Compliance","year":2025,"lang":"","type":"preprint","venue":"SSRN Electronic Journal","topic":"Biomedical Text Mining and Ontologies","field":"Biochemistry, Genetics and Molecular Biology","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"McMaster University","funders":"","keywords":"Reliability (semiconductor); False positive paradox; Domain (mathematical analysis); Similarity (geometry); Proxy (statistics); Scram; Precision and recall","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":[0.02048597,0.001042068,0.001918517,0.004207029,0.002018187,0.004335259,0.002942858,0.004003205,0.005560396],"category_scores_gemma":[0.1811559,0.001021791,0.001388304,0.005044134,0.003641254,0.01705855,0.007515107,0.003497888,0.00482795],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00142835,"about_ca_system_score_gemma":0.004659235,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003822624,"about_ca_topic_score_gemma":0.002764181,"domain_scores_codex":[0.9692314,0.01148709,0.003628044,0.002903551,0.01125641,0.001493571],"domain_scores_gemma":[0.7954401,0.1043567,0.009263822,0.06954527,0.02005212,0.001341952],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001513683,0.0005813584,0.02157503,0.001517272,0.0003376685,0.001370231,0.004836904,0.04285127,0.02690875,0.0747231,0.04448945,0.7792954],"study_design_scores_gemma":[0.0001576521,0.0004575312,0.004870447,0.0004154997,0.000511663,0.00212021,0.002971928,0.6222125,0.08783189,0.248389,0.02981066,0.0002511127],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.230186,0.001959615,0.7258421,0.005710355,0.0003759707,0.0006267867,0.002653856,0.02296552,0.00967978],"genre_scores_gemma":[0.7823106,0.0006443314,0.2016741,0.0009421637,0.0001909876,0.0002726184,0.004249521,0.003841101,0.005874538],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.02048597,"threshold_uncertainty_score":0.1083415,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01565000766946264,"score_gpt":0.2903834509394865,"score_spread":0.2747334432700239,"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."}}