{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.00132858,0.0003849658,0.0005906972,0.0002130041,0.0002548936,0.000499028,0.001422153,0.000209614,0.00001240338],"category_scores_gemma":[0.003207074,0.0002803313,0.0002153476,0.001533679,0.0001243279,0.0008338341,0.0006889859,0.000869882,0.000001502009],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001943538,"about_ca_system_score_gemma":0.0005263403,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001261499,"about_ca_topic_score_gemma":0.00001130381,"domain_scores_codex":[0.9970633,0.0001308855,0.0007093885,0.0009708176,0.0005270311,0.0005985322],"domain_scores_gemma":[0.9977541,0.000314502,0.000469534,0.0003396566,0.001018411,0.0001038227],"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":[0.0000367019,0.00009015189,0.0003310889,0.0006037488,0.00001208854,0.000004311628,0.01025128,0.00001530748,0.07508254,0.003835751,0.0000313815,0.9097056],"study_design_scores_gemma":[0.0007611783,0.00005352912,0.001331054,0.001479248,0.00003923463,0.00004027698,0.002960736,0.8170574,0.1661569,0.009350971,0.0001921178,0.0005772885],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1905611,0.05128775,0.7307347,0.008224932,0.002156371,0.00143702,0.00001465273,0.0003926049,0.01519086],"genre_scores_gemma":[0.6142212,0.000008536137,0.3849525,0.0001995598,0.0001300736,0.0000345247,0.000003005717,0.00001990828,0.0004307117],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9091284,"threshold_uncertainty_score":0.9999649,"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."}}