{"id":"W4405143960","doi":"10.1145/3673791.3698434","title":"The First Workshop on Evaluation Methodologies, Testbeds and Community for Information Access Research (EMTCIR 2024)","year":2024,"lang":"en","type":"article","venue":"","topic":"Advanced Graph Neural Networks","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Computer science; Data science; World Wide Web; Knowledge management","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":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.2213254,0.00341844,0.002872788,0.006530925,0.004348823,0.01816567,0.01017718,0.008895416,0.02106662],"category_scores_gemma":[0.1699941,0.002047935,0.003546625,0.004300762,0.004803279,0.01756812,0.01873156,0.01051805,0.00920385],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0115447,"about_ca_system_score_gemma":0.01753459,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01808162,"about_ca_topic_score_gemma":0.02255527,"domain_scores_codex":[0.8459035,0.1265527,0.005265046,0.005889904,0.01167062,0.00471825],"domain_scores_gemma":[0.8072998,0.09196837,0.00405364,0.01957035,0.04962709,0.02748075],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.001736421,0.002119088,0.002543248,0.002597503,0.000439401,0.0004498818,0.004955642,0.003713956,0.01176534,0.03881676,0.7295907,0.2012722],"study_design_scores_gemma":[0.001503367,0.002307899,0.006587485,0.00287958,0.0004599119,0.0005506735,0.006526351,0.02691617,0.01806289,0.06303591,0.8706427,0.0005271865],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"methods","genre_gemma":"other","genre_scores_codex":[0.0371883,0.01657875,0.657632,0.1083261,0.041461,0.02624492,0.01921095,0.0135478,0.07981011],"genre_scores_gemma":[0.1450753,0.004282529,0.6581193,0.01708445,0.005863479,0.0330481,0.05519791,0.007330377,0.07399855],"genre_candidate":"other","genre_consensus":null,"teacher_disagreement_score":0.2213254,"threshold_uncertainty_score":0.9602444,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.4441747097121654,"score_gpt":0.5203434357644982,"score_spread":0.07616872605233271,"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."}}