{"id":"W3154971029","doi":"10.18653/v1/2021.eacl-main.8","title":"BERxiT: Early Exiting for BERT with Better Fine-Tuning and Extension to Regression","year":2021,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":88,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute; University of Waterloo","funders":"Vector Institute; Natural Sciences and Engineering Research Council of Canada; Government of Canada; Canadian Institute for Advanced Research","keywords":"Computer science; Inference; Fine-tuning; Extension (predicate logic); Acceleration; Code (set theory); Quality (philosophy); Look-ahead; Regression; Machine learning; Artificial intelligence; Algorithm; Programming language","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":[0.0001364654,0.0000858136,0.0001061509,0.00004002815,0.0001261931,0.0001394103,0.0001418313,0.00003130104,0.000006847405],"category_scores_gemma":[0.00004828001,0.00006193618,0.00001807701,0.0001206973,0.000007258165,0.0002428488,0.0002548673,0.00006024509,0.00000320349],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000009777517,"about_ca_system_score_gemma":0.00002205512,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0000218506,"about_ca_topic_score_gemma":0.00002837763,"domain_scores_codex":[0.9991232,0.00001686106,0.0001180581,0.0004107168,0.0001395036,0.0001916772],"domain_scores_gemma":[0.9993542,0.00008465431,0.00002873862,0.0003430034,0.0001076863,0.00008173783],"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":[0.00003656579,0.00004271858,0.01492081,0.00008524794,0.00002500271,0.0001643709,0.00422928,0.0005866858,0.06245617,0.03169989,0.002417956,0.8833353],"study_design_scores_gemma":[0.002323617,0.0005642986,0.04061044,0.001629345,0.0000297659,0.0002870868,0.0003540079,0.8361627,0.09765042,0.008703998,0.01051858,0.001165671],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.3869141,0.00003287802,0.6086661,0.003802619,0.00005210531,0.00007061689,1.876203e-7,0.0000478681,0.0004135176],"genre_scores_gemma":[0.4225998,7.483844e-7,0.5751327,0.001275646,0.00005859385,0.000008570163,7.95883e-7,0.000006498597,0.0009166932],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.8821696,"threshold_uncertainty_score":0.2525684,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0309562390184768,"score_gpt":0.2613005794683832,"score_spread":0.2303443404499064,"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."}}