Nanning Guo's research interests focus on large language models (LLMs), encompassing their design, optimization, and societal impact. He aims to improve LLM efficiency through techniques like model compression, sparse training, and adaptive inference. Nanning also studies alignment, seeking methods to enhance safety, reduce biases, and ensure controllable generation. Another key area of her research is multimodal LLMs, where she explores how language models integrate vision, audio, and structured data. Additionally, she investigates interpretability to understand how LLMs reason and make decisions. She examines ethical challenges, including misinformation and fairness, to guide responsible deployment. Her goal is to advance both the capabilities and trustworthiness of LLMs for real-world applications.