{"id":"W3153557349","doi":"10.1145/3404835.3462951","title":"Not All Relevance Scores are Equal: Efficient Uncertainty and Calibration Modeling for Deep Retrieval Models","year":2021,"lang":"en","type":"article","venue":"","topic":"Information Retrieval and Search Behavior","field":"Computer Science","cited_by":17,"is_retracted":false,"has_abstract":true,"ca_institutions":"Microsoft (Canada)","funders":"","keywords":"Ranking (information retrieval); Relevance (law); Metric (unit); Calibration; Bayesian probability; Divergence-from-randomness model; Process (computing); Scope (computer science); Bayesian inference","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":[0.0003754418,0.0001194017,0.0001482835,0.00006424566,0.0001816444,0.0003947493,0.0002451045,0.00006999625,0.000007307746],"category_scores_gemma":[0.0001816919,0.0001016145,0.00005892127,0.0002646365,0.00002505394,0.000964487,0.0001903928,0.0001031731,0.000003730072],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005228913,"about_ca_system_score_gemma":0.0001125516,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0000185156,"about_ca_topic_score_gemma":0.00001778802,"domain_scores_codex":[0.9986153,0.00003964117,0.000312586,0.0003253355,0.0004146601,0.0002925297],"domain_scores_gemma":[0.9989169,0.0001497393,0.00007305294,0.0002952525,0.0004436554,0.0001214629],"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":[0.00004894916,0.00003407769,0.00001025737,0.00004191638,0.000006313714,0.000004690012,0.000810441,0.8778144,0.001268712,0.1158752,0.00004891304,0.004036143],"study_design_scores_gemma":[0.0003927847,0.00005132107,0.00001296655,0.00002181873,0.000005296965,0.00000704065,0.0001352306,0.9871798,0.007958568,0.004007574,0.00007744224,0.0001501338],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.09923021,0.0001145966,0.8989248,0.001110188,0.0001317651,0.0002505635,0.000008686368,0.0001126345,0.0001165845],"genre_scores_gemma":[0.950639,0.00004097804,0.04796983,0.001074598,0.00003228248,0.0000137504,0.00002201222,0.000007258205,0.0002002923],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8514088,"threshold_uncertainty_score":0.4143717,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06952907239637028,"score_gpt":0.2941728339338712,"score_spread":0.2246437615375009,"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."}}