{"id":"W4287206427","doi":"10.48550/arxiv.2104.09399","title":"TREC Deep Learning Track: Reusable Test Collections in the Large Data\\n Regime","year":2021,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Domain Adaptation and Few-Shot Learning","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Microsoft (Canada)","funders":"","keywords":"Overfitting; Computer science; Reuse; Test set; Set (abstract data type); Deep learning; Track (disk drive); Data set; Information retrieval; Test data; Artificial intelligence; Training set; Artificial neural network; Test (biology); Selection (genetic algorithm); Data mining; Programming language","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":[0.03206664,0.002217854,0.001997288,0.007429505,0.003224721,0.006555925,0.006812853,0.00304726,0.02701928],"category_scores_gemma":[0.09547029,0.001053841,0.002077367,0.00808324,0.002213909,0.006423216,0.007478415,0.006673403,0.02425178],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.005075205,"about_ca_system_score_gemma":0.007141308,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.04651406,"about_ca_topic_score_gemma":0.08431774,"domain_scores_codex":[0.9808286,0.005926576,0.001555316,0.002103337,0.008183085,0.001403179],"domain_scores_gemma":[0.8851414,0.03172327,0.002209475,0.03801581,0.03590038,0.007009594],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"observational","study_design_scores_codex":[0.0003273113,0.0004747477,0.002471381,0.0004469839,0.000179369,0.0001500066,0.0001294455,0.004236732,0.002226678,0.003682894,0.904085,0.08158936],"study_design_scores_gemma":[0.00101578,0.00110394,0.02222944,0.0005284636,0.0002210005,0.0008154953,0.0007388868,0.07781917,0.03084264,0.02313865,0.8411322,0.0004143526],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"dataset","genre_gemma":"empirical","genre_scores_codex":[0.06155149,0.007933231,0.1866424,0.02169791,0.01544092,0.004720582,0.4978056,0.07014174,0.1340661],"genre_scores_gemma":[0.1002445,0.00146491,0.09243716,0.003840159,0.001474003,0.003556601,0.7427948,0.008926866,0.04526097],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.04651406,"threshold_uncertainty_score":0.1695865,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09775210036667754,"score_gpt":0.2110734881241472,"score_spread":0.1133213877574697,"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."}}