{"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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0001446678,0.0001916384,0.0002669175,0.0005381095,0.00035378,0.0002492181,0.0002888165,0.00005958555,0.00003597832],"category_scores_gemma":[0.00000371022,0.0001776596,0.0001273745,0.0008637397,0.00007253693,0.0003125221,0.000007620226,0.0001293075,0.000004516587],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003304805,"about_ca_system_score_gemma":0.00002392249,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001420631,"about_ca_topic_score_gemma":0.000192482,"domain_scores_codex":[0.9986755,0.00007310294,0.0003314509,0.0005808491,0.0001689537,0.0001701887],"domain_scores_gemma":[0.9990759,0.000139678,0.00007253476,0.0003531644,0.0002460995,0.0001125699],"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":[0.00002479138,0.0002064018,0.00003308452,0.00002976861,0.0002378039,9.171763e-7,0.000265276,0.001101879,0.1316525,0.001724495,0.00003506492,0.864688],"study_design_scores_gemma":[0.0000845721,0.0001679483,0.001055364,0.0000169248,0.000144055,0.000002180039,0.00002732879,0.2995324,0.6979558,0.0006890766,0.0001490892,0.0001752495],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.005055995,0.00006701839,0.9930923,0.001111693,0.0001127198,0.0003262385,0.0000687987,0.00009339777,0.00007186155],"genre_scores_gemma":[0.9675403,0.0001446843,0.03077372,0.001231955,0.00001885673,0.00009568581,0.00003392079,0.000005723306,0.0001551794],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9624843,"threshold_uncertainty_score":0.724475,"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."}}