| Aug 09, 2025 | Introducing ASearcher — a 32B search agent trained with AReaL that exhibits extreme long-horizon agentic execution capability. It achieves SOTA performance compared with other open-source models of the same size. Check out our open-sourced code, dataset, and checkpoints! |
| Aug 01, 2025 | We refactored the legacy AReaL code into AReaL-lite, a light-weight and algorithm-first codebase that prioritizes better development experiences for AI researchers. Check out the quickstart guide to begin your journey with AReaL-lite! |
| Jun 23, 2024 | Introducing ReaLHF, a highly efficient system for RLHF training of LLMs. It can achieve up to 10x higher training speedup than existing open-source systems! Check our open-sourced code and the documentation to get started with ReaL quickly! 🚀 |
| Jun 02, 2024 | Our ICML 2024 paper has been selected as an oral presentation! Congrats to Shusheng and see you in Vienna again! |
| May 18, 2024 | Our bipedal motion paper has been short-listed as a finalist for the Best Demo Award at ICRA Expo 2024! Congrats to Yunfei! |
| Apr 29, 2024 | I will be attending ICLR and ICRA 2024. See you in Vienna and Yokohama! 🫵 |