I'm an Algorithm Researcher based in Shanghai.
Previously, I worked at Shanghai AI Laboratory, where I first led the MMPose open-source project before shifting my focus to post-training for the InternLM-series.
I later joined the RL Team at Moonshot.AI as Technical Staff, contributing to the post-training of the Kimi series, including K0, K1, K1.5, K2, and K2.5.
Currently, I work at Xiaohongshu dots Team. As a core contributor to dots3-note, I participated in the 2026 International Mathematical Olympiad(IMO) AI Test, where the agent reached the gold-medal standard and achieved a perfect score.
Some previous research works,
- Recursive Self-Improvement & Long-horizon Reasoning
- IMO2026 Gold Medal(Perfect score)
- VibeSearchBench
- Large Language Models pretraining and alignment
- LLM-based agent applications
- Human-centric perception tasks, focusing on human pose estimation
Blog
- RTMPose (arXiv 2023): [technical report]
- RTMO (CVPR 2024 Poster): [paper]
- RTMW & RTMW3D: [technical report]
