{"id":"W4406610920","doi":"10.1109/tpami.2025.3531907","title":"T2I-CompBench++: An Enhanced and Comprehensive Benchmark for Compositional Text-to-Image Generation","year":2025,"lang":"en","type":"article","venue":"IEEE Transactions on Pattern Analysis and Machine Intelligence","topic":"Image Retrieval and Classification Techniques","field":"Computer Science","cited_by":18,"is_retracted":false,"has_abstract":true,"ca_institutions":"Huawei Technologies (Canada)","funders":"National Natural Science Foundation of China","keywords":"Computer science; Artificial intelligence; Benchmark (surveying); Pattern recognition (psychology); Image (mathematics); Image processing; Image segmentation; Natural language processing; Computer vision; Cartography","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":[0.003482985,0.004398509,0.001463305,0.002603895,0.00101951,0.003370058,0.006782752,0.003696909,0.02252269],"category_scores_gemma":[0.02115714,0.0009515057,0.002063441,0.002430202,0.001107898,0.003831903,0.003064294,0.002412331,0.01355366],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002269319,"about_ca_system_score_gemma":0.002310387,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01496386,"about_ca_topic_score_gemma":0.01678067,"domain_scores_codex":[0.996608,0.000855881,0.0003845563,0.0007010784,0.001187729,0.000262738],"domain_scores_gemma":[0.9910861,0.004660817,0.0002964219,0.001541293,0.001947622,0.0004677406],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.002650083,0.001910041,0.004384985,0.005309531,0.0005996538,0.001440391,0.0005745599,0.1277295,0.02903604,0.009189348,0.3870901,0.4300858],"study_design_scores_gemma":[0.001006813,0.001257248,0.002507938,0.0003256803,0.0001649929,0.001086776,0.0004780229,0.8022717,0.06876104,0.008977127,0.1129546,0.0002080097],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"software","genre_gemma":"methods","genre_scores_codex":[0.130169,0.007445551,0.3342273,0.002239822,0.003271786,0.004425935,0.06960239,0.3801995,0.06841877],"genre_scores_gemma":[0.2429157,0.001800387,0.4516312,0.001780573,0.0003049291,0.002850456,0.2493355,0.02689791,0.02248321],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.02252269,"threshold_uncertainty_score":0.07534587,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02323211433851815,"score_gpt":0.303801744757863,"score_spread":0.2805696304193449,"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."}}