arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2407.02392cs.CV
收藏

TokenPacker:面向多模态LLM的高效视觉投影器

TokenPacker: Efficient Visual Projector for Multimodal LLM

Zhejiang University(浙江大学);Ant Group(蚂蚁集团);NUAA(南京航空航天大学);The Hong Kong Polytechnical University(香港理工大学)

Wentong Li, Yuqian Yuan, Jian Liu, Dongqi Tang, Song Wang, Jie Qin, Jianke Zhu, Lei Zhang

AI总结:

提出TokenPacker视觉投影器,采用由粗到细方案注入多级区域特征生成浓缩视觉token,在压缩75%~89%token的同时保持或提升MLLM性能与效率。

摘要:

视觉投影器是多模态大语言模型(MLLM)中视觉编码器与LLM之间的重要桥梁。通常,MLLM采用简单的MLP通过一对一变换来保留所有视觉上下文。然而,视觉token存在冗余,且在处理高分辨率图像时会大幅增加,显著损害MLLM的效率。近期一些工作引入了resampler或abstractor来减少生成的视觉token数量,但它们无法捕捉更精细的细节,削弱了MLLM的视觉推理能力。在本工作中,我们提出了一种新型视觉投影器,采用由粗到细的方案注入丰富特征以生成浓缩的视觉token。具体而言,我们首先将视觉特征插值为低分辨率的点查询,提供整体视觉表示作为基础。然后,我们引入一个区域到点注入模块,利用高分辨率、多层次的区域线索作为细粒度参考键和值,使其在对应的局部上下文区域内被充分吸收。该步骤有效更新了粗略的点查询,将其转化为丰富的点查询供后续LLM推理使用。大量实验表明,我们的方法压缩了75%~89%的视觉token,同时在多个基准上取得了相当甚至更好的性能,效率显著提升。源代码可在https://github.com/CircleRadon/TokenPacker获取。

英文摘要:

The visual projector serves as an essential bridge between the visual encoder and the Large Language Model (LLM) in a Multimodal LLM (MLLM). Typically, MLLMs adopt a simple MLP to preserve all visual contexts via one-to-one transformation. However, the visual tokens are redundant and can be considerably increased when dealing with high-resolution images, impairing the efficiency of MLLMs significantly. Some recent works have introduced resampler or abstractor to reduce the number of resulting visual tokens. Unfortunately, they fail to capture finer details and undermine the visual reasoning capabilities of MLLMs. In this work, we propose a novel visual projector, which adopts a coarse-to-fine scheme to inject the enriched characteristics to generate the condensed visual tokens. In specific, we first interpolate the visual features as a low-resolution point query, providing the overall visual representation as the foundation. Then, we introduce a region-to-point injection module that utilizes high-resolution, multi-level region-based cues as fine-grained reference keys and values, allowing them to be fully absorbed within the corresponding local context region. This step effectively updates the coarse point query, transforming it into an enriched one for the subsequent LLM reasoning. Extensive experiments demonstrate that our approach compresses the visual tokens by 75%~89%, while achieves comparable or even better performance across diverse benchmarks with significantly higher efficiency. The source codes can be found at https://github.com/CircleRadon/TokenPacker.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑