{"id":"W3120832022","doi":"10.18653/v1/2021.emnlp-main.526","title":"Towards Zero-Shot Knowledge Distillation for Natural Language Processing","year":2021,"lang":"en","type":"article","venue":"Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing","topic":"Topic Modeling","field":"Computer Science","cited_by":25,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"Computer science; Distillation; Task (project management); Benchmark (surveying); Artificial intelligence; Knowledge transfer; Natural language processing; Transfer of learning; Domain knowledge; Variety (cybernetics); Shot (pellet); Machine learning; Domain (mathematical analysis); Knowledge management; Engineering","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":[0.002804106,0.001473253,0.001283544,0.001031002,0.0008584849,0.001870214,0.002591586,0.002300761,0.003460127],"category_scores_gemma":[0.01026699,0.0005637097,0.0009698111,0.001060881,0.002212296,0.005875145,0.005291356,0.005015397,0.001858512],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001154235,"about_ca_system_score_gemma":0.001832829,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003193562,"about_ca_topic_score_gemma":0.005228842,"domain_scores_codex":[0.9982198,0.0006653931,0.00007379331,0.000416614,0.0004626081,0.0001617195],"domain_scores_gemma":[0.9959189,0.002680113,0.0001446402,0.000791066,0.0003164377,0.0001489747],"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":[0.0008926634,0.0005786886,0.001432345,0.0007288549,0.0001647401,0.0003145717,0.0007015468,0.4216012,0.01423633,0.09028196,0.02090286,0.4481641],"study_design_scores_gemma":[0.00003101061,0.00007478667,0.00009023704,0.00002461689,0.00001064329,0.00004874352,0.00005091702,0.9316118,0.005343916,0.06034838,0.002352605,0.0000123803],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.04841625,0.001343383,0.9365972,0.00138038,0.0001463205,0.0001262248,0.000594325,0.006360498,0.005035567],"genre_scores_gemma":[0.563806,0.0007706717,0.4213525,0.001059048,0.0002138718,0.0003093253,0.002960196,0.0008664651,0.008661884],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.003460127,"threshold_uncertainty_score":0.0148297,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08645918045858457,"score_gpt":0.4336183586249349,"score_spread":0.3471591781663503,"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."}}