RL Data ScalingBinfeng Xu
Agent Research · NVIDIA
I focus on advancing post-training and RL systems for Agents – infra, algorithms, harness and data scaling, in a unified and accessible way. I believe that AGI is grokking of collective interests.
For example, I built Polar, the first open framework bridging Agent post-training and harness engineering. Skill2Env democratizes RL data scaling via collective Skills. My recent research covers RL systems and algorithms, terminal/computer-use agent, harness design and continual learning.
Research
RL Data Scaling
Async RL Recipe
Memory Efficient RL
Computer-use Agent
Agent Harness / AutoResearch
Agent RL Infra
Agent RL Infra
Multimodal LLM
RL Recipe
Agent Framework
Agent ReasoningBlog
Misc
I play every game by Hidetaka Miyazaki. I enjoy art. I’m a minimalist. I love cats.
