{"id":"W4385573711","doi":"10.18653/v1/2022.findings-emnlp.385","title":"Continuation KD: Improved Knowledge Distillation through the Lens of Continuation Optimization","year":2022,"lang":"en","type":"article","venue":"","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Continuation; Computer science; Benchmark (surveying); Generalization; Distillation; Artificial intelligence; Limiting; Noise (video); Machine learning; Image (mathematics); Mathematics; Engineering; Programming language","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":[],"consensus_categories":[],"category_scores_codex":[0.0005064516,0.0001012811,0.0001194106,0.00006282913,0.0004576352,0.00006076936,0.0005631336,0.00002838139,0.0001259107],"category_scores_gemma":[0.0001429274,0.00008251058,0.00005561865,0.0006468294,0.00004264072,0.0004268461,0.0002878228,0.0001628081,0.000011275],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00007717306,"about_ca_system_score_gemma":0.00006243162,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000472423,"about_ca_topic_score_gemma":0.00002872285,"domain_scores_codex":[0.9987648,0.0002592219,0.0003416073,0.0002852287,0.0002142544,0.0001349117],"domain_scores_gemma":[0.9986268,0.0002640121,0.0003346592,0.0005369791,0.0002204256,0.00001715676],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00001130918,0.0001536582,0.001063612,0.00000883208,0.00001976645,4.908498e-8,0.004837546,0.6095483,0.003970692,0.3522856,0.0005011414,0.0275995],"study_design_scores_gemma":[0.0003168105,0.00005723212,0.01069672,0.00000218206,0.000009707117,0.000001667049,0.0001277906,0.9823396,0.000348176,0.001410002,0.004592523,0.00009755996],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0108851,0.00004810163,0.9741973,0.003562339,0.000179069,0.0005740392,0.000005997356,0.0001806207,0.01036743],"genre_scores_gemma":[0.9399326,0.000003698373,0.05921784,0.000112838,0.00003666588,0.0001747156,0.00005763302,0.000009675537,0.0004542914],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9290475,"threshold_uncertainty_score":0.3519806,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01687823670323815,"score_gpt":0.274094663465086,"score_spread":0.2572164267618479,"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."}}