Dialogue Systems, Language Generation
The Side Effects of Being Smart: Safety Risks in MLLMs' Multi-Image Reasoning