{"id":"W3103686454","doi":"10.1145/3331184.3331339","title":"The Impact of Score Ties on Repeatability in Document Ranking","year":2019,"lang":"en","type":"preprint","venue":"","topic":"Information Retrieval and Search Behavior","field":"Computer Science","cited_by":19,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Ranking (information retrieval); Search engine indexing; Computer science; Context (archaeology); Information retrieval; Index (typography); Data mining; World Wide Web","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.05446163,0.001120407,0.002184066,0.001605314,0.002916967,0.005476492,0.002877401,0.002222027,0.002498655],"category_scores_gemma":[0.3426875,0.001407139,0.001293721,0.002708544,0.003449292,0.009025191,0.004465922,0.003683061,0.001452129],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002012555,"about_ca_system_score_gemma":0.002560427,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002391397,"about_ca_topic_score_gemma":0.002787053,"domain_scores_codex":[0.880695,0.05787193,0.01160239,0.01369508,0.03232198,0.003813585],"domain_scores_gemma":[0.4643735,0.3425356,0.03115387,0.1356211,0.02220061,0.004115392],"domain_codex":null,"domain_gemma":"reproducibility","domain_candidate":"reproducibility","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.009548764,0.002704492,0.1446417,0.002306722,0.001856231,0.0008501682,0.004932581,0.1292396,0.1299356,0.02312855,0.01758744,0.5332681],"study_design_scores_gemma":[0.0009906025,0.013833,0.1896962,0.0005081517,0.001199543,0.002842846,0.003239658,0.4168828,0.236485,0.09683335,0.0363609,0.001128039],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7455227,0.00263737,0.229727,0.001875437,0.0007457497,0.00115614,0.001181604,0.007303564,0.009850428],"genre_scores_gemma":[0.9329051,0.0002938906,0.06137424,0.0003954693,0.0002049878,0.0005151656,0.001024066,0.001208101,0.002078984],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9455384,"threshold_uncertainty_score":0.2880241,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03803432110190978,"score_gpt":0.3364973686898886,"score_spread":0.2984630475879788,"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."}}