作者位次 #2/13ACL 2026Poster· Natural Language Processing
Are We Using the Right Benchmark: An Evaluation Framework for Visual Token Compression Methods
Chenfei Liao · Wensong Wang · Zichen Wen · Xu Zheng · Yiyu Wang · Haocong He · Yuanhuiyi Lyu · Lutao Jiang · Xin Zou · Yuqian Fu · Bin Ren · Linfeng Zhang · Xuming Hu
摘要
Recent efforts to accelerate inference in Multimodal Large Language Models (MLLMs) have largely focused on visual token compression. The effectiveness of these methods is commonly evaluated by measuring the accuracy drop on existing MLLM benchmarks before and after compression. However, these ben...