How Does Vision-Language Adaptation Impact the Safety of Vision Language Models?

TL;DR

This paper investigates how vision-language adaptation affects the safety of large vision-language models, revealing safety degradation and proposing weight merging as a solution to improve safety without sacrificing helpfulness.

Contribution

The study provides an in-depth analysis of safety impacts during VL adaptation and introduces weight merging to better balance safety and helpfulness in LVLMs.

Findings

Safety degrades during VL adaptation even with safe training data

Safety fine-tuning techniques mitigate risks but cause helpfulness reduction

Weight merging effectively reduces safety degradation while maintaining helpfulness

Abstract

Vision-Language adaptation (VL adaptation) transforms Large Language Models (LLMs) into Large Vision-Language Models (LVLMs) for multimodal tasks, but this process often compromises the inherent safety capabilities embedded in the original LLMs. Despite potential harmfulness due to weakened safety measures, in-depth analysis on the effects of VL adaptation on safety remains under-explored. This study examines how VL adaptation influences safety and evaluates the impact of safety fine-tuning methods. Our analysis reveals that safety degradation occurs during VL adaptation, even when the training data is safe. While safety tuning techniques like supervised fine-tuning with safety datasets or reinforcement learning from human feedback mitigate some risks, they still lead to safety degradation and a reduction in helpfulness due to over-rejection issues. Further analysis of internal model weights suggests that VL adaptation may impact certain safety-related layers, potentially lowering overall safety levels. Additionally, our findings demonstrate that the objectives of VL adaptation and safety tuning are divergent, which often results in their simultaneous application being suboptimal. To address this, we suggest the weight merging approach as an optimal solution effectively reducing safety degradation while maintaining helpfulness. These insights help guide the development of more reliable and secure LVLMs for real-world applications.