{"id":"W3044876097","doi":"10.48550/arxiv.2007.11682","title":"Assessing top-$k$ preferences","year":2020,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Mobile Crowdsensing and Crowdsourcing","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Crowdsourcing; Computer science; Preference; Learning to rank; Measure (data warehouse); Rank (graph theory); Artificial intelligence; Information retrieval; Similarity (geometry); Quality (philosophy); Machine learning; Natural language processing; Data mining; Ranking (information retrieval); Mathematics; Statistics; World Wide Web","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0002141641,0.0003274113,0.0003650175,0.0001566428,0.0002388713,0.0008997309,0.001878135,0.0002755242,0.00002093806],"category_scores_gemma":[0.00005786402,0.0003710301,0.0002052522,0.0005408312,0.00009437121,0.0006487925,0.002388224,0.0007715247,0.0001047555],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001178204,"about_ca_system_score_gemma":0.0003108755,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001165907,"about_ca_topic_score_gemma":0.00001163098,"domain_scores_codex":[0.9977667,0.0001773688,0.0002030626,0.001359619,0.0001181323,0.0003751449],"domain_scores_gemma":[0.9981103,0.0001182721,0.0002523442,0.001167679,0.0001202492,0.0002311425],"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":[0.0000437187,0.0002978302,0.01497197,0.0007887058,0.0004810587,0.003340883,0.003905835,0.4074657,0.001578378,0.5108037,0.003703961,0.05261827],"study_design_scores_gemma":[0.0003319679,0.0000545287,0.002409304,0.0003332385,0.00008584523,0.00001296781,0.0003859596,0.8886201,0.001054872,0.1043012,0.001493853,0.0009161308],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3664495,0.00005554671,0.6234685,0.0002398486,0.0006622802,0.0001258913,0.000002377043,0.0004995279,0.008496553],"genre_scores_gemma":[0.9944445,0.0000385749,0.004708274,0.0001745527,0.0001521178,4.551317e-7,0.000007326907,0.00001699734,0.0004572178],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.627995,"threshold_uncertainty_score":0.9998742,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1363594842486464,"score_gpt":0.2113050918641414,"score_spread":0.07494560761549501,"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."}}