FuzzingRL: 强化模糊测试揭示视觉语言模型的缺陷

Vision Language Models (VLMs) are susceptible to errors, making it crucial to identify where these errors occur to ensure the reliability and safety of AI systems. This research proposes an approach that automatically generates questions designed to deliberately induce incorrect responses from VLMs, thereby revealing their vulnerabilities. The core of this approach lies in fuzz testing and reinforcement fine-tuning: a single input query is transformed into a large set of diverse variants through vision and language fuzzing. Based on the outcomes of fuzz testing, the question generator is further guided by adversarial reinforcement fine-tuning to produce increasingly challenging queries that trigger model failures. This method allows for the consistent identification and analysis of potential flaws in VLMs.