Reproduce “BT-Agent: Deploying Onboard Robot Agents with Behavior Trees and Multi-Fidelity Evaluation” 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
Fabio Hübel, Adam Lilja, Yue Yin, Siming He, Claire Tomlin, Shankar Sastry, Jonas Frey, Marco Pavone
BT-Agent lets a language model choose among predefined robot skills while local behavior trees check whether a running skill may continue and interrupt it without another model call. Its 9B Qwen3.5 NVFP4 configuration reaches 54.7% task-weighted success on text tasks, versus 49.7% for ReAct. Robot trials show commanded velocity reaches zero a median 19 ms after local acceptance, not physical stopping.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.