About LaMI Lab: Building Smarter, More Reliable AI Agents Our research group is committed to building AI agents that are practical and trustworthy. Toward this end, we are currently exploring the following topics: Advancing the capabilities of (Multimodal) Large Language Models: We're pushing the boundaries of what large language models and multimodal models can do, making them fundamentally more capable and versatile. Furthermore, we are interested in broadening the scope of research to (M)LLM-driven physical AI agents, such as VLA systems, to be deployed in the real world. Connecting (M)LLMs to real-world data: We are working on seemlessly augmenting the knowledge base of these models to integrate and learn from diverse real-world data, including graphs, time-series, and tables. Ensuring fair and trustworthy evaluation: We are exploring new evaluation frameworks to ensure that these advanced AI agents are reliable and transparent, no matter the application.