Yongbin Li
Head of the Qoder & Tongyi Lingma Models; Algorithm Scientist at Tongyi Lab
Li Yongbin graduated from Tsinghua University and serves as Head of the Qoder and Tongyi Lingma models, as well as an Algorithm Scientist at Tongyi Lab. His research focuses on large model technologies, including post-training, code models, character models, and AI agents. In the AI coding domain, he led the development of the industry-leading Qwen-Coder-Qoder large model and helped build products such as Qoder and Tongyi Lingma. He is also responsible for large model development across multiple application areas, including AI social companionship (Qwen-Character, Tongyi Stardust), AI customer service (Intelligent Customer Service, Tongyi Xiaomi), and AI-powered data analysis. In recent years, he has published over 100 papers at top international conferences, including ICLR, NeurIPS, ICML, ACL, EMNLP, and CVPR. His work has received more than 6,500 citations on Google Scholar, with an H-index of 37. He has also served as an Area Chair for leading international conferences such as ACL, NAACL, and WSDM.
Topic
QwenStar Character Model - Explore modeling a “person”!
By modeling the world's knowledge, the Big Language Model demonstrates powerful information processing, understanding and reasoning capabilities. The Character Model goes one step further, aiming to explore the complex entity of “human being” in depth, and attempts to portray human knowledge, emotion, personality, memory, morality and mental activities in multiple dimensions. Based on the Character Model, a variety of applications can be developed, including role-playing, real-life replicas, emotional accompaniment, and digital people, which can be widely used in various scenarios, such as social networking, gaming, and consumer electronics, to provide users with a more personalized and immersive interactive experience. QwenStar is committed to in-depth research and innovative exploration of character models, with the goal of building “interesting and useful” multilingual and multimodal Humanoid Agents. This presentation will systematically introduce QwenStar's latest progress in the field of Character Model, including the overall architecture, core technology, application cases and future development direction.
