{"id":"W4416017238","doi":"10.1145/3746252.3760922","title":"LLM-as-a-Judge in Entity Retrieval: Assessing Explicit and Implicit Relevance","year":2025,"lang":"","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto; Toronto Metropolitan University","funders":"","keywords":"Relevance (law); Context (archaeology); Replicate; Benchmark (surveying); Reliability (semiconductor)","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01706973,0.002047289,0.001534125,0.004166787,0.001225343,0.003850087,0.002295613,0.003286047,0.004150207],"category_scores_gemma":[0.08118684,0.0006139582,0.001146301,0.002315905,0.001097682,0.006736964,0.004172982,0.002879898,0.004481783],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001374272,"about_ca_system_score_gemma":0.001763471,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01133637,"about_ca_topic_score_gemma":0.02121259,"domain_scores_codex":[0.9880232,0.006680356,0.0008008144,0.002477732,0.001490205,0.0005278484],"domain_scores_gemma":[0.9543294,0.03225468,0.001996611,0.006381601,0.003597761,0.001439877],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.008020369,0.001870191,0.202702,0.00384004,0.001449146,0.0008981912,0.005524239,0.0775982,0.02717453,0.01384271,0.1199189,0.5371615],"study_design_scores_gemma":[0.000340853,0.0007049069,0.03346821,0.000243865,0.0001923614,0.0004254408,0.001258313,0.902324,0.01865566,0.0169545,0.0251646,0.0002672696],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.602371,0.008370657,0.2863081,0.002687008,0.001015041,0.0012545,0.01909197,0.05769538,0.02120631],"genre_scores_gemma":[0.8092378,0.0004259901,0.1602592,0.000708606,0.0002235385,0.0004664299,0.02294186,0.001531864,0.004204551],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01706973,"threshold_uncertainty_score":0.09027445,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02380253752750885,"score_gpt":0.3127308103459749,"score_spread":0.2889282728184661,"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."}}