Multimodal Safety Evaluation in Generative Agent Social Simulations

摘要

Can generative agents be trusted in multimodal environments? Despite recent advances, agents remain limited in their ability to reason about safety, coherence, and trust across modalities. We introduce a reproducible simulation framework to evaluate generative agents in three aspects: (1) safety improvement over time via iterative plan revision in multimodal scenarios; (2) detection of unsafe activities across social contexts; and (3) social dynamics, measured through interaction and acceptance rates. These multimodal agents are evaluated using metrics that quantify plan revisions and unsafe-to-safe conversions. Experiments show that while agents detect direct multimodal contradictions, they often fail to align local revisions with global safety, achieving only a 55% success rate in correcting unsafe plans. We release a dataset of 1,000 multimodal plans, yielding more than 600,000 simulation steps. Notably, 45% of unsafe actions are accepted when paired with misleading visual cues, revealing a strong tendency to overtrust visual content. Code is available at https://github.com/AdonaiVera/X-CASE