On the morning of July 17, 2025, Yang Jianchao, the head of ByteDance's Doubao large model's vision multimodal generation division, announced an "internal leave" within the company, and his responsibilities have been transitioned accordingly. Currently, Yang Jianchao's information remains accessible within ByteDance's internal system.

According to company insiders, Yang Jianchao's duties will be assumed by Zhou Chang, alias "Shiguang," who currently serves in the "Multimodal Interaction and World Model" department and reports to Wu Yonghui.

The reason behind this personnel change is attributed to "family factors," insiders revealed.

Rumors also circulated that Yang Jianchao had experienced physical and mental exhaustion due to the challenge of balancing his workload between North America and China under prolonged high-pressure conditions. Another version suggested that he was opting for an "early retirement."

This announcement likely comes on the heels of the completion of mid-year performance evaluations, providing an opportune moment for work realignment in the second half of the year, as noted by insiders.

Yang Jianchao has a long-standing career in the vision field. After securing the Guo Moruo Scholarship at the University of Science and Technology of China in 2006, he studied under Thomas Huang, known as "the father of computer vision," and completed his postdoctoral research at the University of Illinois at Urbana-Champaign. He has held positions at Adobe and Snapchat, specializing in visual algorithms. In 2018, he joined ByteDance's AI Lab as a research and development director, later taking charge of the Seed Vision department in 2023.

Taking over the reins, Zhou Chang is a prominent figure in the domestic technology sector, holding a bachelor's degree from Fudan University and a Ph.D. from Peking University. He previously led the technical research and development of Alibaba's Tongyi Qianwen and spearheaded the creation of the M6 multimodal model, released in 2021—the largest Chinese AI model jointly developed by Alibaba and Tsinghua University, marking a significant milestone in Alibaba's large model strategy. In July 2025, Zhou Chang left Alibaba to join ByteDance's Seed team.

The Seed Vision Team maintains a global footprint with offices in San Jose (North America), Singapore, and various locations in China, focusing on image-text-video generation and fundamental visual research. Since Wu Yonghui, a former executive of Google's DeepMind, joined and took charge of Seed's fundamental research in February 2025, the team has undergone several rounds of organizational restructuring.

Amidst the current fierce competition in large models, the change in leadership has sparked concerns about the stability of ByteDance's AI roadmap. However, insiders emphasized that ByteDance has consistently reaffirmed its unwavering commitment to fundamental research.