{"id":"W4389520096","doi":"10.18653/v1/2023.findings-emnlp.500","title":"Co-training and Co-distillation for Quality Improvement and Compression of Language Models","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Kootenay Association for Science & Technology","funders":"","keywords":"Computer science; Margin (machine learning); Benchmark (surveying); Distillation; Inference; Language model; Performance improvement; Machine learning; Artificial intelligence; Performance prediction; Simulation; 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":[0.003124317,0.002068764,0.001644991,0.001317347,0.0008850262,0.001647451,0.002956016,0.001786893,0.003320401],"category_scores_gemma":[0.01296693,0.0008130927,0.001486503,0.001580354,0.001435997,0.004394765,0.003624813,0.005183468,0.001770528],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001222173,"about_ca_system_score_gemma":0.002469998,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.009330746,"about_ca_topic_score_gemma":0.01487955,"domain_scores_codex":[0.9978421,0.0007422282,0.0001335438,0.0006438991,0.0004125061,0.0002257046],"domain_scores_gemma":[0.9953068,0.002619538,0.0002069777,0.001231617,0.0004778155,0.0001573555],"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":[0.0006575108,0.0003737241,0.002586877,0.0004158661,0.0003002309,0.0003231951,0.0004130351,0.3621013,0.01955182,0.01562398,0.0120619,0.5855906],"study_design_scores_gemma":[0.00003094871,0.00006824964,0.0002263335,0.0000171901,0.00003014524,0.00006220272,0.00002816775,0.9812882,0.009033582,0.007297598,0.001895744,0.00002163722],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.04991543,0.001881156,0.9317831,0.0008447697,0.0002647258,0.0001368684,0.0004537952,0.01156451,0.003155738],"genre_scores_gemma":[0.6003213,0.0007212887,0.3891706,0.0007759141,0.000201908,0.0002952732,0.002575408,0.001056453,0.004881822],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.009330746,"threshold_uncertainty_score":0.0185529,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1113164359390328,"score_gpt":0.3746026075313697,"score_spread":0.2632861715923369,"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."}}