{"id":"W4393335160","doi":"10.48550/arxiv.2403.18910","title":"A Geometric Explanation of the Likelihood OOD Detection Paradox","year":2024,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Industrial Vision Systems and Defect Detection","field":"Engineering","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Canadian Institute for Advanced Research","keywords":"Maximum likelihood; Econometrics; Computer science; Mathematical economics; Statistics; Mathematics","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":[],"consensus_categories":[],"category_scores_codex":[0.0002288529,0.0002070863,0.0002438642,0.000599896,0.00006828465,0.00003952344,0.000252477,0.0004395248,0.0000214289],"category_scores_gemma":[0.00003801956,0.0001871715,0.0002714447,0.001549392,0.00002386954,0.00006244778,0.0002644736,0.0007166055,0.0000683153],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002776666,"about_ca_system_score_gemma":0.00004657475,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002190404,"about_ca_topic_score_gemma":0.00009603926,"domain_scores_codex":[0.9990554,0.0000704548,0.0002463633,0.0003419981,0.0001093039,0.0001764353],"domain_scores_gemma":[0.9992241,0.00005219067,0.000126575,0.0004616689,0.00008417614,0.00005128416],"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":[0.00005859813,0.00004497087,0.0004004237,0.0009299387,0.0003979997,0.0000355808,0.0002393173,0.976971,0.003804212,0.001438701,0.0009090977,0.01477019],"study_design_scores_gemma":[0.001419334,0.0002001545,0.004068851,0.001242392,0.0008110495,0.00003037353,0.000583117,0.8840932,0.07622342,0.02622207,0.003917941,0.001188109],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9610259,0.0003055262,0.03047602,0.000006320121,0.004565871,0.0004167314,0.00003517946,0.0003234253,0.002845064],"genre_scores_gemma":[0.9993831,0.00007517674,0.000005236202,0.000002764887,0.000206227,0.00000216778,0.000004135276,0.00003117333,0.0002900297],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.09287778,"threshold_uncertainty_score":0.7632634,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04486886149411207,"score_gpt":0.1648855912950577,"score_spread":0.1200167298009457,"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."}}