Liang Wenfeng, founder of Chinese AI startup DeepSeek, shared the best paper award at the Association for Computational Linguistics (ACL) conference in Vienna, Austria. ACL is one of the most important AI research events in the world.
The winning paper, called “Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention,” was published in February. Liang was one of 15 authors. The paper explains a new method that helps DeepSeek’s AI models work faster and cheaper.
This award highlights how Chinese researchers are leading in AI language studies. More than half of the main authors at this year’s ACL conference came from China, up from less than 30% last year. The US came second with 14%.
Two of the four best papers at the conference were from China. One was from Liang’s team at DeepSeek, and the other was from Yang Yaodong’s team at Peking University.
Liang keeps a low profile, and it is not clear if he attended the conference in person. In May, he co-wrote another paper about how DeepSeek built a powerful open-source AI system at a lower cost than others.
DeepSeek is focusing on research rather than quick fundraising or business moves. However, its model use in China dropped from 99% to 80% recently, showing strong competition in the AI field.
