Vibe Coding DiscoverASearcher
View on GitHubAn Open-Source Large-Scale Reinforcement Learning Project for Search Agents
Open-source code, data pipelines, and training guidance for building web-search agents with large-scale asynchronous reinforcement learning. Includes QA synthesis, search tools, and evaluation workflows.
Use Cases
Train web-search agents with reinforcement learningBuild long-horizon agents that use search toolsSynthesize challenging grounded QA training dataEvaluate search agents on GAIA, xBench-DeepSearch, and Frames
Built With
- Language
- Python
- Frameworks
- AReaL · SGLang · Ray
Tags
search agent · reinforcement learning · agent training · web search · tool use · asynchronous RL · QA synthesis · evaluation