{"id":"W3104738015","doi":"10.18653/v1/2020.sustainlp-1.11","title":"Early Exiting BERT for Efficient Document Ranking","year":2020,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":49,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo; Vector Institute","funders":"Natural Sciences and Engineering Research Council of Canada; Canada First Research Excellence Fund","keywords":"Computer science; Speedup; Inference; Ranking (information retrieval); Latency (audio); Computation; Language model; Code (set theory); Artificial intelligence; Parallel computing; Algorithm; Programming language; Set (abstract data type); Telecommunications","routes":{"ca_aff":true,"ca_fund":true,"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":[0.001729615,0.001701503,0.001254863,0.001236173,0.0008935836,0.001884429,0.002155864,0.001226848,0.01249713],"category_scores_gemma":[0.00813763,0.000862199,0.0009243182,0.001204585,0.0005944972,0.003650278,0.001952584,0.002650719,0.01065849],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009775879,"about_ca_system_score_gemma":0.002234764,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.009610074,"about_ca_topic_score_gemma":0.02638284,"domain_scores_codex":[0.998768,0.0003593814,0.0000722957,0.0002418334,0.0003680541,0.0001904286],"domain_scores_gemma":[0.9971389,0.001311861,0.0001272621,0.0008388365,0.0004236075,0.0001595436],"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":[0.001356525,0.0003248618,0.003676387,0.000459322,0.0001388254,0.000294949,0.0003560502,0.1042892,0.02399333,0.0173819,0.05034155,0.7973871],"study_design_scores_gemma":[0.00007296451,0.0001470048,0.00052284,0.00002138053,0.00003101273,0.0001614376,0.00005796837,0.9593586,0.0121224,0.01564727,0.01181099,0.0000460922],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.02266447,0.001111207,0.9044222,0.0003437098,0.0002227686,0.0002228935,0.001303355,0.0661926,0.00351674],"genre_scores_gemma":[0.3408755,0.0005193771,0.6324785,0.000514139,0.0002933759,0.0003453282,0.006277509,0.003852718,0.01484353],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.01249713,"threshold_uncertainty_score":0.04180706,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03341623580851892,"score_gpt":0.2571902084438234,"score_spread":0.2237739726353045,"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."}}