Recently, a major personnel update has emerged in the industry: Yonglong Tian, a former researcher at OpenAI, has joined Tencent and is expected to serve as the head of the multimodal model direction of Tencent Hunyuan, leading the development of vision-language models (VLMs). Although this change has not been officially announced yet, the news has already attracted widespread attention in the industry. If this collaboration is successfully finalized, Tian will become another core research talent to join the Tencent Hunyuan team, following Shunyu Yao.

Tian's joining marks Tencent's continued investment in the field of multimodal large models. He and Tencent's current Chief AI Scientist Shunyu Yao once worked together at OpenAI, witnessing the evolution of the GPT series and multimodal technologies. As an outstanding talent in computer vision, Tian has a solid academic background: he graduated from Tsinghua University, completed his master's degree at the MMLab of The Chinese University of Hong Kong, and then pursued his doctorate at the Massachusetts Institute of Technology (MIT). His research trajectory clearly demonstrates a transition from traditional visual recognition to generative models and representation learning, with nearly 30,000 citations on Google Scholar.

image.png

In academia, Tian has made significant contributions. His proposal of "Supervised Contrastive Learning" extended the contrastive learning framework to supervised scenarios, profoundly influencing the design concepts of visual pre-training models. Additionally, his recent research focuses on autoregressive image generation, particularly the Fluid series, by verifying the superiority of continuous tokens in visual quality, providing new technical pathways for large model generation capabilities. Before joining Tencent, he worked at Google Research and Google DeepMind, and later joined OpenAI as a technical member.

This personnel introduction is an important part of the recent restructuring and talent strategy of the Tencent Hunyuan multimodal team. Since Tencent officially established the Large Language Model Department and the Multimodal Model Department in April 2025, the team has been active. Previously, several experienced talents from research institutions such as Microsoft, Alibaba, and Tsinghua University have joined, aiming to address past issues of "dispersed resources" and build a more competitive multimodal R&D system.

With the addition of tech giants like Tian, Tencent Hunyuan's research capabilities in multimodal and visual perception are expected to further improve. As a strong competitor in the top tier of the AI industry, Tencent's series of talent "puzzle" actions indicate its deep layout and strategic ambitions in the multimodal capabilities of the Hunyuan model.