Calling a Spade a Heart: Gaslighting Multimodal Large Language Models via Negation

Multimodal Large Language Models (MLLMs) have exhibited remarkable advancements in integrating different modalities, excelling in complex understanding and generation tasks. Despite their success, MLLMs remain vulnerable to conversational adversarial inputs, particularly negation arguments. This paper systematically evaluates state-of-the-art MLLMs across diverse benchmarks, revealing significant performance drops when negation arguments are introduced to initially correct responses. We show critical vulnerabilities in the reasoning and alignment mechanisms of these models. Proprietary models such as GPT-4o and Claude-3.5-Sonnet demonstrate better resilience compared to open-source counterparts like Qwen2-VL and LLaVA. However, all evaluated MLLMs struggle to maintain logical consistency under negation arguments during conversation. This paper aims to offer valuable insights for improving the robustness of MLLMs against adversarial inputs, contributing to the development of more reliable and trustworthy multimodal AI systems.