Baidu Wenxin is rumored to have two key members: North American Frontier Lab+DeepSeek OCR core author Dynamic Beating AI News: Baidu's Basic Model Development Department (BMU) is strengthening its core model team. According to insiders, the person in charge, Sun Tianxiang, introduced a researcher named "Wu Qin" from a Frontier Lab in North America to work with the team to strengthen the pre training of Wenxin. The other party has participated in the development of several generations of cutting-edge models in the past two years, but their real name and original affiliation have not been disclosed. On the other hand, former DeepSeek researcher Wei Haoran and his team have also transferred from within Baidu to BMU, where he serves as the head of Wenxin Multimodal Algorithm. Wei Haoran's line actually had a foreshadowing long ago. When Baidu opened up Unlimited OCR in June, the technical director among the three core contributors only signed "YY", and the community speculated at the time that it was Wei Haoran. Wei Haoran is the first author of DeepSeek OCR and DeepSeek OCR 2; The DeepSeeker V4 technical report also marked him as having resigned. When Unlimited OCR was released, it achieved 93.92% on OmniDocBench v1.6, breaking the end-to-end OCR record at one point. It adopts DeepSeek OCR's DeepEncoder and redesigns the attention mechanism of the decoding end, which can continuously parse dozens of pages of documents at once. Sun Tianxiang took over BMU in July and joined the Baidu Model Committee. At the end of the month, he joined the Baidu Technology Strategy Committee. Two months later, both the pre training and multimodal core lines of Wenxin began to be supplemented. [Original link]