MoVE-KD: Knowledge Distillation for VLMs with Mixture of Visual Encoders

TL;DR

MoVE-KD introduces a novel knowledge distillation framework that efficiently combines multiple visual encoders into a single model, leveraging input-dependent specialization and attention mechanisms to improve vision-language model performance.

Contribution

The paper proposes MoVE-KD, a new method that distills multiple visual encoders into one using low-rank adaptation, mixture-of-experts, and attention-based distillation strategies.

Findings

Effective in consolidating multiple encoders into a single model.

Improves performance on VLM benchmarks like LLaVA and LLaVA-NeXT.

Reduces computational cost while maintaining high accuracy.

Abstract

Visual encoders are fundamental components in vision-language models (VLMs), each showcasing unique strengths derived from various pre-trained visual foundation models. To leverage the various capabilities of these encoders, recent studies incorporate multiple encoders within a single VLM, leading to a considerable increase in computational cost. In this paper, we present Mixture-of-Visual-Encoder Knowledge Distillation (MoVE-KD), a novel framework that distills the unique proficiencies of multiple vision encoders into a single, efficient encoder model. Specifically, to mitigate conflicts and retain the unique characteristics of each teacher encoder, we employ low-rank adaptation (LoRA) and mixture-of-experts (MoEs) to selectively activate specialized knowledge based on input features, enhancing both adaptability and efficiency. To regularize the KD process and enhance performance, we propose an attention-based distillation strategy that adaptively weighs the different encoders and emphasizes valuable visual tokens, reducing the burden of replicating comprehensive but distinct features from multiple teachers. Comprehensive experiments on popular VLMs, such as LLaVA and LLaVA-NeXT, validate the effectiveness of our method. Our code is available at: https://github.com/hey-cjj/MoVE-KD.