{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.001799382,0.0005878996,0.0009414927,0.0004224985,0.0003559634,0.0005336948,0.002389537,0.0004361887,0.00007243126],"category_scores_gemma":[0.0001935797,0.0007849633,0.0002570774,0.000886665,0.0001961061,0.001362145,0.005636149,0.0006287398,0.00000408908],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004505861,"about_ca_system_score_gemma":0.0006310242,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0009087165,"about_ca_topic_score_gemma":0.0002624976,"domain_scores_codex":[0.995203,0.0003818666,0.0007355872,0.00271482,0.0002256368,0.0007390502],"domain_scores_gemma":[0.9965143,0.0002979986,0.0005862133,0.001669083,0.0004773656,0.0004551042],"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":[0.000454035,0.0004180313,0.005534335,0.001058899,0.0006891366,0.0006253887,0.04134459,0.5994327,0.004351392,0.2940288,0.00002685801,0.05203585],"study_design_scores_gemma":[0.003307792,0.0003016182,0.003212122,0.001836066,0.000223813,0.00002985431,0.006048261,0.9683118,0.002216361,0.01239572,0.0007686598,0.001347887],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3691848,0.0001279823,0.6287074,0.000169961,0.0006159872,0.0009159056,0.00001405993,0.00003889261,0.0002250362],"genre_scores_gemma":[0.8443405,0.0002273124,0.1540246,0.0001157087,0.00007604525,0.000004673497,0.00001435396,0.00003437718,0.001162381],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.4751557,"threshold_uncertainty_score":0.9994601,"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."}}