{"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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.000169645,0.00006817575,0.0000842442,0.0000194044,0.00008283372,0.0001473233,0.0004262103,0.00001922257,0.00001451269],"category_scores_gemma":[0.00003923522,0.00006036475,0.00005093157,0.00009986715,0.000005309056,0.0001125651,0.0001920759,0.00004704021,0.00002719193],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001895148,"about_ca_system_score_gemma":0.00001931543,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002139287,"about_ca_topic_score_gemma":6.180672e-7,"domain_scores_codex":[0.9991546,0.00001157752,0.0001609823,0.0002958429,0.0001699073,0.0002071164],"domain_scores_gemma":[0.9995992,0.00006554873,0.00003378235,0.0001913376,0.00003156829,0.00007861617],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00001117053,0.00002564485,0.0003528423,0.00006018394,0.00002315583,0.000008449209,0.009170636,0.03363331,0.002500816,0.757014,0.0005699483,0.1966298],"study_design_scores_gemma":[0.0003333665,0.00004177606,0.00008381939,0.00001047834,0.000002244723,0.000001025368,0.00003466802,0.9924619,0.002620579,0.001687454,0.002616404,0.0001063147],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.06366643,0.00002461229,0.9273529,0.005346505,0.0001525361,0.0001859077,1.692493e-7,0.0001865602,0.003084397],"genre_scores_gemma":[0.7464905,2.627459e-7,0.2518408,0.001444097,0.0001040318,0.00001284199,1.928894e-7,0.000004099433,0.0001032378],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9588286,"threshold_uncertainty_score":0.2461603,"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."}}