{"id":"W4389518890","doi":"10.18653/v1/2023.findings-emnlp.634","title":"Mixture-of-Linguistic-Experts Adapters for Improving and Interpreting Pre-trained Language Models","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of the Fraser Valley; University of British Columbia","funders":"","keywords":"Softmax function; Computer science; Language model; Gumbel distribution; Adapter (computing); Layer (electronics); Artificial intelligence; Pruning; Natural language processing; Encoding (memory); Artificial neural network; Mathematics; Statistics; Extreme value theory","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":[0.003376855,0.002568076,0.001185117,0.001727093,0.0006285995,0.001623832,0.003299846,0.002188223,0.005494905],"category_scores_gemma":[0.009425808,0.001208838,0.001818099,0.001184651,0.001033624,0.004487673,0.003405413,0.004232154,0.003293649],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00115788,"about_ca_system_score_gemma":0.001388437,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005494606,"about_ca_topic_score_gemma":0.01164022,"domain_scores_codex":[0.9986634,0.000437797,0.00007617758,0.0004400801,0.0002383454,0.0001441124],"domain_scores_gemma":[0.9973236,0.001330319,0.0001690362,0.0005596176,0.0004795377,0.0001379553],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0006455531,0.0002583055,0.003234278,0.0002403225,0.0004118868,0.000311361,0.0007986975,0.2185383,0.02478127,0.0145453,0.006855562,0.7293791],"study_design_scores_gemma":[0.00002214615,0.00006089004,0.000394577,0.00002395237,0.00007318716,0.0000887423,0.00005561815,0.9771491,0.009653242,0.0096168,0.002836193,0.00002560413],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01522173,0.0003943198,0.9773933,0.0001728816,0.00006293035,0.00007706029,0.00008490783,0.005271866,0.001320931],"genre_scores_gemma":[0.3756603,0.0005081143,0.6117528,0.0006193133,0.0001885834,0.0003833288,0.001123652,0.001077865,0.00868614],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.005494905,"threshold_uncertainty_score":0.01838225,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02013140030204983,"score_gpt":0.2764942223730839,"score_spread":0.2563628220710341,"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."}}