Junlin Xie and co-authors investigate the burgeoning field of large multimodal agents (LMAs) in Large Multimodal Agents: A Survey. The paper presents a comprehensive review of AI-powered LLM agents, mapping their expansion into the multimodal domain, and categorization of existing research. It also proposes a unified framework for evaluating LMAs to foster meaningful comparisons. The possibilities and future research directions are structured to help guide novices in this evolving landscape.

• Surveys the development of LMAs that extend beyond text-based applications
• Categorizes LMAs into four types and collaborative frameworks
• Suggests a standard framework for the evaluation of LLMs
• Identifies applications and potential research trajectories in multimodal AI

Stimulating further research in LMAs, this comprehensive study serves as a gateway for those embarking on multimodal agent ventures and facilitates the development of more adept AI across various platforms.