I'm an algorithm researcher based in Shanghai, currently working at Xiaohongshu dots Team. As a core contributor, I participated in the International Mathematical Olympiad(IMO) 2026 AI Test, where our dots-note-3.0 model achieved a perfect-score gold-medal standard.
Previously, I was a Technical Staff on the RL Team at Moonshot.AI, where I contributed to the post-training of the Kimi series (K0, K1, K1.5, K2, K2.5).
Before that, I was at Shanghai AI Laboratory, working on post-training for the InternLM-series and leading the MMPose open-source project.
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]
