{"id":"W4389523905","doi":"10.18653/v1/2023.emnlp-main.83","title":"How Does Generative Retrieval Scale to Millions of Passages?","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":23,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Computer science; Generative grammar; Transformer; Information retrieval; Search engine indexing; Document retrieval; Artificial intelligence; Encoder; Generative model; Ranking (information retrieval); Task (project management); Natural language processing","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":[0.0001566767,0.00005770476,0.00009303126,0.0001041578,0.0000487131,0.00007282293,0.0004060882,0.00002797202,0.00001146332],"category_scores_gemma":[0.00005223378,0.00003911516,0.00003496479,0.0006507757,0.0000135283,0.0001643324,0.0002707333,0.0000420636,0.00003971028],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001226772,"about_ca_system_score_gemma":0.00002414258,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0000176529,"about_ca_topic_score_gemma":0.00003665135,"domain_scores_codex":[0.9992799,0.00002635685,0.00009987662,0.0002336768,0.0002030442,0.0001571887],"domain_scores_gemma":[0.9994062,0.00004901366,0.00002337163,0.0003852127,0.00006816622,0.00006801886],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0000213782,0.0001437904,0.00598627,0.00008134193,0.0001019069,0.00006318487,0.01856751,0.01040433,0.4286698,0.364994,0.06004401,0.1109225],"study_design_scores_gemma":[0.0003057671,0.0001219308,0.00565595,0.00002963419,0.000005909606,0.000002867725,0.000595998,0.4424432,0.5298735,0.01199291,0.008629372,0.0003428996],"study_design_candidate":"bench_or_experimental","study_design_consensus":"bench_or_experimental","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1736668,0.000006654788,0.8090999,0.01564391,0.0002808773,0.00008778015,0.000002703214,0.0001561586,0.001055252],"genre_scores_gemma":[0.8117418,0.000006450355,0.1732855,0.0003182053,0.00007929641,0.000004539959,8.925208e-7,0.000004411954,0.01455892],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.638075,"threshold_uncertainty_score":0.159507,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03023236006595731,"score_gpt":0.2627108354550887,"score_spread":0.2324784753891314,"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."}}