[DeepSeek Releases New Paper by Liang Wenfeng: Proposing the mHC Architecture to Enhance Stability in Large Model Training] According to Jin10, DeepSeek has released a new paper proposing a novel architecture called Manifold-Constrained Hyperconnection (mHC). This architecture aims to address issues in Hyperconnection (HC) technology, such as training instability and limited scalability caused by the disruption of identity mapping properties. By mapping the residual connection space of HC to a specific manifold to restore identity mapping properties, and combining it with rigorous infrastructure optimization to ensure efficiency, the architecture achieves significant performance improvements and superior scalability. DeepSeek anticipates that mHC, as a flexible and practical extension of HC, will contribute to a deeper understanding of topological architecture design and point to promising directions for the evolution of foundational models. The paper lists Zhenda Xie, Yixuan Wei, and Huanqi Cao as co-first authors, with Liang Wenfeng also included among the authors.