I’m a graduate student at Nanjing University of Aeronautics and Astronautics, exploring LLM agents, agentic reinforcement learning, and the algorithms behind large language models.
I follow my curiosity through papers and small experiments. On my blog, I share reading notes and questions I’m still thinking through.
- Qwen2.5 Post-Training · Training scripts and notes for SFT, DPO, PPO, and GRPO, with evaluation code for GSM8K.
- Build GPT-2 · A from-scratch implementation of GPT-2 covering pretraining, fine-tuning, and text generation.
- Reinforcement Learning on CartPole · Implementations of DQN, policy gradient, actor–critic, and PPO on CartPole.

