Zhang, Y. et al. 2026. LLaVA-UHD v2: Exploiting Hierarchical Vision Granularity in MLLMs via Inverse Semantic Pyramid. Proceedings of the AAAI Conference on Artificial Intelligence. 40, 15 (Mar. 2026), 12934–12942. DOI:https://doi.org/10.1609/aaai.v40i15.38292.