{"id":"W4414197327","doi":"10.1109/cvprw67362.2025.00056","title":"Revisiting Referring Expression Comprehension Evaluation in the Era of Large Multimodal Models","year":2025,"lang":"en","type":"article","venue":"","topic":"Speech and dialogue systems","field":"Computer Science","cited_by":9,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Benchmark (surveying); Comprehension; Vocabulary; Object (grammar); Benchmarking; Key (lock)","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":[0.001607332,0.00006210363,0.0001139813,0.00007861995,0.00006549,0.00004061932,0.0003996891,0.00003998643,0.000004392221],"category_scores_gemma":[0.00007211779,0.0000403397,0.00003330668,0.0003204783,0.000006324606,0.0003061858,0.0001177226,0.0001050639,0.000003139222],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002702572,"about_ca_system_score_gemma":0.00004662497,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001601163,"about_ca_topic_score_gemma":0.00002013012,"domain_scores_codex":[0.9987901,0.0002931088,0.0002579443,0.0001925891,0.0003293785,0.000136818],"domain_scores_gemma":[0.9992669,0.0001623654,0.00006559458,0.0003921567,0.00009944678,0.00001348838],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00008000401,0.0003755313,0.008548999,0.0003656939,0.00002103738,0.0000110957,0.01255029,0.06428677,0.1617122,0.4474562,0.001884507,0.3027077],"study_design_scores_gemma":[0.0005260846,0.000008351849,0.006275498,0.0003253435,0.000002148435,8.768915e-7,0.000271721,0.9797961,0.007860624,0.004834509,0.00005079631,0.00004799986],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3254526,0.0003534975,0.6500186,0.0004877097,0.0002247716,0.000439699,7.186662e-7,0.00004883897,0.02297355],"genre_scores_gemma":[0.9863986,0.000002983076,0.01332404,0.0002140084,0.00002398136,0.00001471486,0.000002974233,0.000001600431,0.000017051],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9155093,"threshold_uncertainty_score":0.1645005,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04613760366182853,"score_gpt":0.3210189740561321,"score_spread":0.2748813703943035,"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."}}