VoCo-LLaMA: Towards Vision Compression with Large Language Models
Authors:
Top Cited Papers
- 10 June 2025
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 29836-29846
Abstract
Vision-Language Models (VLMs) have achieved remarkable success in various multi-modal tasks, but they are often bottlenecked by the limited context window and high computational cost of processing high-resolution image inputs and videos. Vision compression can alleviate this problem by reducing the vision token count. Previous approaches compress vision tokens with external modules and force LLMs to understand the compressed ones, leading to visual information loss. However, the LLMs’ understanding paradigm of vision tokens is not fully utilised in the compression learning process. We propose VoCo-LLaMA, the first approach to compress vision tokens using LLMs. By introducing Vision Compression tokens during the vision instruction tuning phase and leveraging attention distillation, our method distill how LLMs comprehend vision tokens into their processing of VoCo tokens. VoCo-LLaMA facilitates effective vision compression and improves the computational efficiency during the inference stage. Specifically, our method can achieve a 576× compression rate while maintaining 83.7% performance. Furthermore, through continuous training using time-series compressed token sequences of video frames, VoCo-LLaMA demonstrates the ability to understand temporal correlations, outperforming previous methods on popular video question-answering benchmarks. Our approach presents a promising way to unlock the full potential of VLMs’ contextual window, enabling more scalable multi-modal applications. The project page can be accessed via https://yxxxb.github.io/VoCo-LLaMA-page.Keywords
All Related Versions
This publication has 25 references indexed in Scilit:
- An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Models2024
- MMBench: Is Your Multi-modal Model an All-Around Player?2024
- LLaMA-VID: An Image is Worth 2 Tokens in Large Language Models2024
- Segment and Caption Anything2024
- BT-Adapter: Video Conversation is Feasible Without Video Instruction Tuning2024
- Visual Programming: Compositional visual reasoning without training2023
- EVA: Exploring the Limits of Masked Visual Representation Learning at Scale2023
- GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering2019
- Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering2017
- MSR-VTT: A Large Video Description Dataset for Bridging Video and Language2016