Han, Shuhao, et al. “EvalMuse-40K: A Fine-Grained Benchmark With Comprehensive Human Annotations for Text-to-Image Generation Model Alignment Evaluation”. Proceedings of the AAAI Conference on Artificial Intelligence, vol. 40, no. 6, Mar. 2026, pp. 4583-91, doi:10.1609/aaai.v40i6.42458.