Liang Wenfeng 梁文锋

Positions

• Founder · DeepSeek

Selected publications

• DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning — arXiv, 2025

Companies

• High-Flyer (幻方量化) — co-founder — quant fund funding DeepSeek's compute

Collaborators & adjacent figures

• M 87 · Yang Zhilin the two primary names in Chinese open-weight reasoning models

Observation notes

Funded frontier research from a quant fund's GPU stockpile and released open-weight models at a fraction of assumed training cost. Primary name (with Yang Zhilin) for distillation-related tracking topics.