2 papers
cs.RO2026
NavHarness: Towards Lifelong Embodied Navigation
Xunyi Zhao, Jian Zhou, Sihao Lin +6
Frontier models can now perform well on individual embodied navigation tasks through multi-round multimodal reasoning with simple tools. Across successive tasks, however, an agent…
cs.AI2026
CUA-Sandbox: Efficient Environments for Computer-Use Agent Reinforcement Learning
Xin Yan, Zhengbo Jiao, Jiaqi Liu +11
Reinforcement learning enables computer-use agents to improve through interaction with real software environments, including websites and desktop applications. However, conventiona…