ByteDance's big model division, Seed, completed a new round of organizational adjustments last week. The most critical part is that Seed Foundation Model has established four new first-level divisions: Pretrain Data, headed by Lee Sung-gang. Pre-training data teams previously distributed in text models, programming, visual understanding, and speech were merged to unify the multi-modal data for the new Omni model and the data work required for pre-training the oversized model. Horizon RL, head Tang Shengyu. It integrates post-training teams previously scattered around post-training, reasoning, and visual understanding, and is mainly responsible for strengthening learning and raising the upper limit of the model's basic intelligence. Product Posttrain-Work, head of Qin Yujia. The new department mainly serves B-side applications, is responsible for the integration and release of Agentic models, and optimizes Agentic capabilities for office scenarios, including task models that support Doubao and Dola. The Application Team, which was originally headed by Zhu Wenjia, continues to be retained and renamed Product Posttrain-Chat, which mainly serves C-side applications, integrates and publishes dialogue models, and collaborates with Product Posttrain-Work. The heads of the four new departments all reported to Wu Yonghui. Additionally, Seed has appointed separate heads for AI security and other cutting-edge exploration directions.

Disclaimer:Webull uses external vendor Google Translation Service for news translations where we endeavour to ensure these are correct, however, we recommend that you please double-check this information accordingly. Webull is not responsible for translation errors or issues.