Yuhang Liu is a researcher at Zhejiang University specialising in large language models (LLMs), multimodal agents, and reasoning-enhanced AI. His work focuses on advancing LLM applications, reasoning enhancement, agent-based interactions, and Retrieval-Augmented Generation (RAG). Yuhang developed InfiGUI-R1, a multimodal GUI agent that utilizes reinforcement learning, subgoal-guidance, and error-recovery mechanisms to enable planning and reflection. This innovation allowed a 3B parameter model to match or even surpass the performance of 7B/72B models, achieving state-of-the-art (SOTA) results at its release. Additionally, he designed InfiGUIAgent, a multimodal GUI agent with native reasoning and reflection capabilities, where a 2B parameter model also achieved SOTA performance. Furthering his contributions, Yuhang proposed a novel RAG method that incorporates fine-grained guidance from LLMs, significantly improving retriever performance across five general benchmarks. His work bridges advanced theory with practical applications, driving progress in AI-powered reasoning and multimodal systems.