{"id":"W4226000139","doi":"10.1093/imaiai/iaae002","title":"PACMAN: PAC-style bounds accounting for the Mismatch between Accuracy and Negative log-loss","year":2024,"lang":"en","type":"preprint","venue":"Information and Inference A Journal of the IMA","topic":"Machine Learning and Data Classification","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"École de Technologie Supérieure; McGill University","funders":"Universidad de Buenos Aires; Consejo Nacional de Investigaciones Científicas y Técnicas","keywords":"Generalization; Computer science; Algorithm; Function (biology); Metric (unit); Generalization error; Cross entropy; Mathematics; Artificial intelligence; Artificial neural network; Pattern recognition (psychology)","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":["scholarly_communication"],"consensus_categories":[],"category_scores_codex":[0.001286353,0.0001596465,0.0002115266,0.000124863,0.0003378356,0.002083702,0.0009936307,0.0001071833,0.000003250942],"category_scores_gemma":[0.001477176,0.00008496188,0.00009527427,0.0001485661,0.0001014622,0.001650916,0.001187219,0.0008733403,0.000006018836],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003629671,"about_ca_system_score_gemma":0.0002621178,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00004521676,"about_ca_topic_score_gemma":0.000004284682,"domain_scores_codex":[0.9987597,0.00007255931,0.00061494,0.0001148085,0.0002953219,0.0001426276],"domain_scores_gemma":[0.9969359,0.001072571,0.001199843,0.0003833447,0.0003548706,0.0000534479],"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":[0.00002806834,0.00001198426,0.008036128,0.0005512098,0.0001897581,7.000118e-7,0.01736304,0.0003704412,0.00001988299,0.02033482,0.004864736,0.9482293],"study_design_scores_gemma":[0.001440247,0.0002430698,0.2171773,0.0023122,0.0003431424,0.0002448506,0.002304819,0.4369344,0.0003852311,0.2010885,0.1367736,0.0007526881],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1925717,0.001878648,0.7051216,0.09500757,0.002455176,0.001087841,0.000163574,0.00009786532,0.001616039],"genre_scores_gemma":[0.9952027,0.0004816519,0.003523265,0.00052718,0.0001971411,0.00001228003,0.000009889551,0.000004668977,0.00004127691],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9474765,"threshold_uncertainty_score":0.9989522,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02670544377808226,"score_gpt":0.3117600080436237,"score_spread":0.2850545642655414,"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."}}