arxivcs.CV2026-07-22
Look Less, Think Faster: Joint Token-Compute Adaptation for Multimodal LLMs
Pengcheng Wang, Zhiquan Wang, Jayoung Lee, Zhuoyan Xu, Ran Xu, Saurabh Bagchi, et al.
Multimodal Large Language Models (MLLMs) have recently demonstrated strong performance across vision-language tasks. However, their high inference cost, arising from both the large number of input visual tokens and the heavy computation of the large language model (LLM), remains…