GDT-VLM: Global Distribution Modeling for Visual Token Compression in Efficient Multimodal Large Language Models.

@inproceedings{xie-2026-gdtvlm, author = {Jiangtao Xie and Junjie Wu and Zhaolin Zhang and Qilong Wang and Peihua Li}, editor = {Stevan Rudinac and Zi Huang and Marcel Worring and Lucia Vadicamo and Xirong Li and Pascal Mettes and Lizi Liao}, title = {{GDT-VLM: Global Distribution Modeling for Visual Token Compression in Efficient Multimodal Large Language Models}}, booktitle = {{Proceedings of the 2026 International Conference on Multimedia Retrieval, ICMR 2026, Amsterdam, The Netherlands, June 16-19, 2026}}, pages = {1711--1715}, publisher = {ACM}, year = {2026}, url = {https://doi.org/10.1145/3805622.3810846}, doi = {https://doi.org/10.1145/3805622.3810846} }