Retrieval-augmented-Generation
Pre-trained neural language models can learn a substantial amount of in-depth knowledge from data without any access to an external memory, as a param
Front Page
Mars Dingdang
Learning in public. Notes on code, systems, and building.
Hi there
I'm Sitian Ding, a freshman undergraduate at IIIS, Tsinghua University.
I’m interested in reinforcement learning, multi-agent systems, LLM post-training, AI agents, and high-performance algorithms / systems. I enjoy building things that connect algorithms, learning systems, and real-world engineering.
About Me
- Undergraduate student at Institute for Interdisciplinary Information Sciences (IIIS), Tsinghua University
- Interested in RL, LLM training systems, multi-agent learning, AI agents, and parallel / high-performance computing
- Experienced with Python, C++, PyTorch, basic CUDA programming, neural network training, LLM post-training, RL self-play, and parallel algorithms
- Former chemistry olympiad contestant, National Gold Medalist
- Former informatics olympiad contestant, NOI / Winter Camp Bronze Medalist
Current Interests
- Reinforcement learning for language models and agents
- Multi-agent learning and self-play
- LLM post-training systems
- AI agents for games and reasoning tasks
- High-performance algorithms and parallel computing
Recent Posts 4 posts Expand Collapse
Transformer-from-scratch
This is adapted from harvard AnnotatedTransformer, which gives an annotated implementation of the Transformer model from the paper “Attention is All Y
my-first-blog
Hello World!
Hello World
Welcome to Hexo! This is your very first post. Check documentation for more info. If you get any problems when using Hexo, you can find the answer in