news
| Nov 19, 2025 | I’ll be at NeurIPS this year; come say hi! |
|---|---|
| Nov 10, 2025 | Scaling up RL with verifiable environments yields surprisingly strong performance. Check out our preprint! |
| Sep 26, 2025 | Contrary to many recent claims, we show that RL actually learns new skills and generalizes surprisingly well in LLMs. Check out our new preprint. |