This is a project dedicated to systematically investigating how to build strong foundation models for Robot-Use Agents (RUA).
Research Question 1:
First, we need to recognize that objective and accurate benchmarking of RUA is the foundation for everything else.
- Revisiting RUA analyzes existing RUA systems and proposes a common protocol for evaluating them.
- uni-robot-agent provides a unified and high-performance harness for RUAs. Designed for generalizable agentic training.
More is in the works. Stay tuned!