Reproduce “Know When to Escalate: Critic-Guided Adaptive Execution for Cloud-Local Embodied Agents” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Wei Zhang, Lirong Xiang, Charles Fleming
CARE coordinates local and cloud executors during embodied tasks, using a critic to escalate when local actions appear unreliable and shared memory to preserve verified progress across handoffs. On Habitat rearrangement, it achieves 92.5% success versus AdaPlanner’s 74.2%, with median time between actions of 0.80 seconds versus Cloud-only’s 4.61. This is a simulation result; CARE scores below Cloud-only on EmbodiedBench.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.