{"id":"W3184918446","doi":"10.18653/v1/2021.repl4nlp-1.17","title":"In-Batch Negatives for Knowledge Distillation with Tightly-Coupled Teachers for Dense Retrieval","year":2021,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":119,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Canada First Research Excellence Fund","keywords":"Encoder; Computer science; Distillation; Ranking (information retrieval); Artificial intelligence; Knowledge transfer; Learning to rank; Information retrieval; Machine learning; Knowledge management","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0002187316,0.00009231047,0.0001334602,0.00004536903,0.00006478161,0.00007733953,0.0001855639,0.00004702047,0.00001062107],"category_scores_gemma":[0.0001893564,0.00007439562,0.00004171859,0.0002529874,0.00001755996,0.000275889,0.00005196531,0.00005400243,0.000003127509],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00006190186,"about_ca_system_score_gemma":0.0001623394,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001049677,"about_ca_topic_score_gemma":0.0002282125,"domain_scores_codex":[0.9991534,0.00002626914,0.0001672007,0.0003600762,0.00009672987,0.0001963716],"domain_scores_gemma":[0.9990944,0.0003675069,0.00004183453,0.000272114,0.0001775728,0.00004654774],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0007098158,0.0006737591,0.0110031,0.0004199174,0.0001943551,0.00004267992,0.05583158,0.01046526,0.02023163,0.8107468,0.003821989,0.08585905],"study_design_scores_gemma":[0.00128386,0.00009921118,0.0009838806,0.00002411518,0.000006822085,0.000004233178,0.0002146254,0.9779423,0.007560244,0.008930424,0.002780983,0.000169323],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1529965,0.00005580904,0.8443881,0.001229093,0.0001492429,0.000296627,0.000001054169,0.0000504945,0.0008331078],"genre_scores_gemma":[0.734412,0.000001624285,0.2608715,0.00008346749,0.00008229586,0.00002806267,0.000003715044,0.000008209892,0.004509179],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.967477,"threshold_uncertainty_score":0.3033765,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02914580604038951,"score_gpt":0.2862546506800759,"score_spread":0.2571088446396864,"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."}}