$WATERWATER ราคา

On August 19, according to Lookonchain, two whale addresses shorted a total of 50,838 ETH on Hyperliquid, with a notional size of approximately $98 million.

On August 19, according to Lookonchain, two whale addresses shorted a total of 50,838 ETH on Hyperliquid, with a nominal scale of approximately $98 million.

According to Beating AI news, ByteDance's large model team Seed has completed a new round of organizational adjustments. The Seed Foundation Model has established four primary departments, merging previously dispersed data and post-training teams across text, code, vision, and speech. The core change is the unification of pre-training data and reinforcement learning, while separating the application post-training into two lines: Work and Chat. The Pretrain Data department is responsible for the multimodal data of the Omni model and the data required for pre-training large models. Horizon RL focuses on reinforcement learning to enhance the model's foundational capabilities. The Product Posttrain-Work department targets office and B-end applications, emphasizing the optimization of model invocation tools, computer operations, and long task execution capabilities. The original Application team has been renamed Product Posttrain-Chat, continuing to manage C-end dialogue models. All four departments report to Wu Yonghui. Previously, Seed was organized by text, code, vision, and speech, with each having its own data and post-training personnel. Now, ByteDance aims for the next generation of models to directly adopt the Omni approach, allowing different modalities to converge on a single foundation. Seed has also been discussing training models with over 50 trillion parameters. As models grow larger and modalities increase, the previously fragmented R&D approach is more prone to redundant investments. Similar adjustments have occurred at Tencent, which merged its mixed Yuan large language model and multimodal teams in July, with Yao Shunyu overseeing them. ByteDance has also specifically established a Work post-training department, indicating that the office Agent has become a distinct product line for optimization by the foundational model team.