METEOR: Multi-Encoder Collaborative Token Pruning for Efficient Vision Language Models

TL;DR

METEOR introduces a multi-stage token pruning framework for multi-encoder vision-language models, significantly reducing computational load while maintaining high performance across diverse benchmarks.

Contribution

It is the first to propose a multi-stage token pruning approach for multi-encoder vision-language models, enhancing efficiency without sacrificing accuracy.

Findings

Reduces 76% visual tokens with only 0.3% performance drop

Achieves state-of-the-art efficiency on 11 benchmarks

Demonstrates effectiveness of collaborative token pruning strategies

Abstract

Vision encoders serve as the cornerstone of multimodal understanding. Single-encoder architectures like CLIP exhibit inherent constraints in generalizing across diverse multimodal tasks, while recent multi-encoder fusion methods introduce prohibitive computational overhead to achieve superior performance using complementary visual representations from multiple vision encoders. To address this, we propose a progressive pruning framework, namely Multi-Encoder collaboraTivE tOken pRuning (METEOR), that eliminates redundant visual tokens across the encoding, fusion, and decoding stages for multi-encoder MLLMs. For multi-vision encoding, we discard redundant tokens within each encoder via a rank guided collaborative token assignment strategy. Subsequently, for multi-vision fusion, we combine the visual features from different encoders while reducing cross-encoder redundancy with cooperative pruning. Finally, we propose an adaptive token pruning method in the LLM decoding stage to further discard irrelevant tokens based on the text prompts with dynamically adjusting pruning ratios for specific task demands. To our best knowledge, this is the first successful attempt that achieves an efficient multi-encoder based vision language model with multi-stage pruning strategies. Extensive experiments on 11 benchmarks demonstrate the effectiveness of our proposed approach. Compared with EAGLE, a typical multi-encoder MLLMs, METEOR reduces 76% visual tokens with only 0.3% performance drop in average. The code is available at https://github.com/YuchenLiu98/METEOR.