The reporter learned that Jiukun Investment and Microsoft Research Asia recently reported that DeepSeek-R1 successfully reproduced for the first time, especially its achievements in the field of strengthening learning, and put forward innovative ideas on the technical level. The academic article is titled Logic RL: Unleashing LLM Reasoning with Rule-Based Reinforcement Learning， Co authored by Microsoft Research Asia, Ubiquant and other independent researchers. It is reported that as a quantitative giant, Jiukun Investment has strong technical and talent reserves in data, algorithms, and computing power. In recent years, it has continued to conduct systematic and in-depth research in the field of AI cutting-edge technology, not only exploring general technologies for a long time and promoting their scenario applications, but also conducting diversified research and expansion in multiple sub fields, striving to build a more comprehensive AI technology system. In the industry’s view, DeepSeek’s models are all open source, sparking a global craze for replicating DeepSeeker R1. Previously, institutions such as HuggingFace, Berkeley University, and Shanghai Jiao Tong University have conducted relevant replication research and achieved certain results. (Reporter Yan Jun from Caixin News Agency)
Scan code to share