CoVFT: Context-aware Visual Fine-tuning for Multimodal Large Language Models
This addresses a fundamental design problem in MLLMs for researchers and practitioners, offering a stable and effective fine-tuning method that is incremental but impactful.
The paper tackles the inconsistency in whether to fine-tune or freeze vision encoders in multimodal large language models (MLLMs), finding that existing methods often underperform frozen baselines due to visual preference conflicts, and proposes CoVFT, a context-aware fine-tuning framework that achieves state-of-the-art performance on 12 benchmarks, with a 7B model outperforming the average of a 13B counterpart.
Multimodal large language models (MLLMs) achieve remarkable progress in cross-modal perception and reasoning, yet a fundamental question remains unresolved: should the vision encoder be fine-tuned or frozen? Despite the success of models such as LLaVA and Qwen-VL, inconsistent design choices and heterogeneous training setups hinder a unified understanding of visual fine-tuning (VFT) in MLLMs. Through a configuration-aligned benchmark, we find that existing VFT methods fail to consistently outperform the frozen baseline across multimodal tasks. Our analysis suggests that this instability arises from visual preference conflicts, where the context-agnostic nature of vision encoders induces divergent parameter updates under diverse multimodal context. To address this issue, we propose the Context-aware Visual Fine-tuning (CoVFT) framework, which explicitly incorporates multimodal context into visual adaptation. By integrating a Context Vector Extraction (CVE) and a Contextual Mixture-of-Experts (CoMoE) module, CoVFT decomposes conflicting optimization signals and enables stable, context-sensitive visual updates. Extensive experiments on 12 multimodal benchmarks demonstrate that CoVFT achieves state-of-the-art performance with superior stability. Notably, fine-tuning a 7B MLLM with CoVFT surpasses the average performance of its 13B counterpart, revealing substantial untapped potential in visual encoder optimization within MLLMs.