{"id":"W4287122359","doi":"10.48550/arxiv.2106.05346","title":"End-to-End Training of Multi-Document Reader and Retriever for\\n Open-Domain Question Answering","year":2021,"lang":"","type":"preprint","venue":"arXiv (Cornell University)","topic":"Topic Modeling","field":"Computer Science","cited_by":45,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Computer science; Question answering; Benchmark (surveying); Information retrieval; Domain (mathematical analysis); Set (abstract data type); Training set; Open domain; Artificial intelligence; Labrador Retriever; Relevance (law); Machine learning; Mathematics","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002474232,0.001968516,0.001672245,0.001023807,0.0006772,0.001374598,0.003932685,0.003679361,0.01084727],"category_scores_gemma":[0.007251043,0.0007951137,0.00131881,0.0008066526,0.0008201555,0.003298829,0.002570907,0.003238576,0.008533395],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001056159,"about_ca_system_score_gemma":0.001347,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003823397,"about_ca_topic_score_gemma":0.008648984,"domain_scores_codex":[0.9984072,0.0004431805,0.00009636638,0.0006693205,0.0002389998,0.0001449116],"domain_scores_gemma":[0.9970692,0.00160571,0.0001384263,0.0005495591,0.0004976548,0.00013941],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0005950548,0.0005593201,0.001719082,0.0004465101,0.000143061,0.0003052012,0.0004756982,0.09038835,0.02476457,0.004237144,0.01867626,0.8576897],"study_design_scores_gemma":[0.00006165457,0.000189288,0.0004402245,0.00002063502,0.00004495204,0.0001685487,0.00008408885,0.9748231,0.0143971,0.004674574,0.005073273,0.00002260177],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.01845829,0.0006531227,0.9610558,0.000344897,0.0000897112,0.0002166611,0.000374613,0.01677416,0.002032734],"genre_scores_gemma":[0.2355661,0.0003314502,0.7431706,0.0008318265,0.0001924636,0.0005735313,0.004147909,0.001012949,0.01417323],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.01084727,"threshold_uncertainty_score":0.03628772,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1484820626929869,"score_gpt":0.244233744139887,"score_spread":0.0957516814469001,"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."}}