{"id":"W4362419701","doi":"10.31237/osf.io/gz5rw","title":"Extended Abstract: CNN-based Patch Benchmarking","year":2023,"lang":"en","type":"preprint","venue":"","topic":"Advanced Neural Network Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Computer science; Artificial intelligence; Benchmark (surveying); Benchmarking; ENCODE; Task (project management); Variety (cybernetics); Image (mathematics); Object (grammar); Matching (statistics); Computer vision; Subroutine; Pattern recognition (psychology); Similarity (geometry); Range (aeronautics); Function (biology); Mathematics","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0002301667,0.000327595,0.0002845616,0.0001581465,0.0001652275,0.0002420281,0.002210788,0.0002297246,0.00008823373],"category_scores_gemma":[0.00002786311,0.0003221934,0.0001709311,0.0004895112,0.00004531731,0.0001481614,0.001895209,0.0007477126,0.0004920056],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009521916,"about_ca_system_score_gemma":0.0001829553,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00009873104,"about_ca_topic_score_gemma":0.0000601215,"domain_scores_codex":[0.99748,0.00002692092,0.000434515,0.001184938,0.0004173304,0.0004562542],"domain_scores_gemma":[0.9970268,0.000393774,0.0002681108,0.002056,0.000107257,0.0001480413],"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":[0.00000977691,0.0003258304,0.0007745934,0.0002943401,0.00008705223,0.0001808337,0.0002415977,0.3150384,0.0008147176,0.1226323,0.04238369,0.517217],"study_design_scores_gemma":[0.0001859827,0.00002274527,0.01602989,0.0001366015,0.00001178635,0.000002684982,0.000005183596,0.8135583,0.001519735,0.1604891,0.007286059,0.0007518933],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.001043338,0.00004920243,0.9858047,0.003935676,0.001177096,0.000552099,0.000009523942,0.002000231,0.005428162],"genre_scores_gemma":[0.5021091,0.00006286058,0.4933047,0.001100708,0.0005329953,0.0005697007,0.0001090777,0.00006552926,0.002145346],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.5164651,"threshold_uncertainty_score":0.999923,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05589014290483632,"score_gpt":0.3102263510594575,"score_spread":0.2543362081546211,"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."}}