xlang-ai/OSWorld
3,135 534PythonApache-2.0Updated 2 months ago
[NeurIPS 2024] OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments
agentartificial-intelligencebenchmarkclicode-generationguilanguage-modellarge-action-modelllmmultimodalnatural-language-processingreinforcement-learningrpavlm
Our verdict: worth borrowing ideas from
Offers a comprehensive simulated desktop environment and evaluation framework for multimodal agents, a valuable reference for building and testing the engineer's Claude‑based agent orchestration.