{"id":"W3214038065","doi":"10.18653/v1/2021.findings-emnlp.26","title":"Multi-Task Dense Retrieval via Model Uncertainty Fusion for Open-Domain Question Answering","year":2021,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":9,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"Natural Sciences and Engineering Research Council of Canada; Canada First Research Excellence Fund; Compute Canada","keywords":"Computer science; Benchmark (surveying); Task (project management); Question answering; Information retrieval; Joint (building); Set (abstract data type); Domain (mathematical analysis); Aggregate (composite); Artificial intelligence; Training set; Code (set theory); Language model; Natural language processing","routes":{"ca_aff":true,"ca_fund":true,"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":[0.0006005304,0.0001275915,0.000161919,0.00004320709,0.0001810483,0.0002599392,0.0007320082,0.00008163628,0.000009102044],"category_scores_gemma":[0.0001094716,0.0001224371,0.00006172563,0.0002012427,0.00001258834,0.0005393943,0.0007678447,0.00009765638,0.00000862739],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001069144,"about_ca_system_score_gemma":0.0001675476,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001162541,"about_ca_topic_score_gemma":0.0001673732,"domain_scores_codex":[0.9986202,0.00006507187,0.0002565943,0.000577475,0.0002025266,0.000278104],"domain_scores_gemma":[0.9989501,0.0000682212,0.00005753628,0.0006220706,0.0002043345,0.00009778755],"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":[0.00008589048,0.0002081639,0.000104146,0.0000651919,0.00002379543,0.00004780904,0.001583391,0.5633613,0.28842,0.08619612,0.0003430864,0.05956105],"study_design_scores_gemma":[0.0007226072,0.00002524754,0.0000338149,0.00002852505,0.000003661491,0.00001597153,0.0000326066,0.9762189,0.008332753,0.0140717,0.0003451907,0.0001690177],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.03650377,0.00005755447,0.9614165,0.0009441503,0.0003003027,0.0003202572,0.000002567857,0.0001374054,0.0003175195],"genre_scores_gemma":[0.3213884,0.000008541575,0.6770898,0.0004038656,0.00003793872,0.00001205035,0.000006173452,0.000008425832,0.001044812],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.4128575,"threshold_uncertainty_score":0.4992839,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04196707278900472,"score_gpt":0.3043780166640182,"score_spread":0.2624109438750135,"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."}}