{"id":"W4407953544","doi":"10.1145/3701551.3705706","title":"LLM4Eval@WSDM 2025: Large Language Model for Evaluation in Information Retrieval","year":2025,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"Microsoft (Canada); University of Waterloo","funders":"Universitas Brawijaya","keywords":"Computer science; Information retrieval; Artificial intelligence","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":[0.04383498,0.002575113,0.002375802,0.002649663,0.001446652,0.006603014,0.005416617,0.004763175,0.03240196],"category_scores_gemma":[0.0707552,0.00135404,0.002961422,0.00166261,0.001710059,0.008037996,0.007154211,0.006804872,0.01370943],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.004279639,"about_ca_system_score_gemma":0.004154136,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0110145,"about_ca_topic_score_gemma":0.01367045,"domain_scores_codex":[0.9662855,0.02575935,0.001318355,0.002028647,0.003917597,0.0006904278],"domain_scores_gemma":[0.953994,0.02860343,0.0007883149,0.009260568,0.004837465,0.002516081],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001877534,0.001162967,0.001609083,0.001758483,0.0007209309,0.0003220015,0.0006906898,0.02445286,0.008119121,0.03898517,0.5406101,0.379691],"study_design_scores_gemma":[0.001374424,0.001481432,0.002981992,0.0004880333,0.0002518907,0.000382488,0.0004264417,0.536203,0.02321251,0.1011309,0.3317084,0.0003585092],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01708268,0.003801386,0.8492559,0.009135071,0.003511323,0.003663533,0.01588195,0.07534406,0.02232411],"genre_scores_gemma":[0.139764,0.001282889,0.7463675,0.003887936,0.0009919198,0.006493067,0.05845746,0.01618494,0.02657037],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.04383498,"threshold_uncertainty_score":0.2318243,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02675071938737008,"score_gpt":0.3243063935755636,"score_spread":0.2975556741881935,"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."}}