{"id":"W4389523983","doi":"10.18653/v1/2023.emnlp-main.362","title":"Fast and Robust Early-Exiting Framework for Autoregressive Language Models with Synchronized Parallel Decoding","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":13,"is_retracted":false,"has_abstract":true,"ca_institutions":"Kootenay Association for Science & Technology","funders":"Institute for Information and Communications Technology Promotion; Ministry of Science and ICT, South Korea; Korea Advanced Institute of Science and Technology","keywords":"Computer science; Security token; Decoding methods; Autoregressive model; Language model; Inference; Estimator; Algorithm; Artificial intelligence; 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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.004048137,0.001642889,0.001769546,0.0009887561,0.0006137304,0.001706663,0.003421019,0.001581562,0.003104839],"category_scores_gemma":[0.01192372,0.001083245,0.001122926,0.000840738,0.001103796,0.003241043,0.002733766,0.00359953,0.001541381],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00123694,"about_ca_system_score_gemma":0.003243371,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008069312,"about_ca_topic_score_gemma":0.01018845,"domain_scores_codex":[0.9980865,0.0007013743,0.0001369274,0.0004491151,0.0004204273,0.0002056636],"domain_scores_gemma":[0.9960917,0.002369642,0.0002452347,0.0005406658,0.0005737563,0.000179014],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0006320337,0.0002031745,0.001644568,0.0001826247,0.0001732732,0.0002757461,0.0004275528,0.5909691,0.01633236,0.05384443,0.00391916,0.331396],"study_design_scores_gemma":[0.00001367582,0.00002223915,0.00005765685,0.000004070278,0.000008769007,0.00001878593,0.000006737899,0.9896666,0.001791153,0.008034283,0.0003654497,0.00001053472],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.003617602,0.0001173612,0.994275,0.00006196868,0.00001812164,0.00002652367,0.000040215,0.001603261,0.0002399897],"genre_scores_gemma":[0.275182,0.0002740571,0.7194163,0.0002137742,0.0001173781,0.0002887572,0.000598691,0.0009164712,0.002992572],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.008069312,"threshold_uncertainty_score":0.02140886,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03884542175441516,"score_gpt":0.2738687050378683,"score_spread":0.2350232832834531,"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."}}