uni-robot-agent introduced to democratize Robot-use Agents research
A unified harness, uni-robot-agent, has been introduced to support research into Robot-use Agents using open-weight models.
Robot-use Agents (RUAs) have shown promising zero-shot manipulation capabilities, but reliance on proprietary frontier models has limited research into further training. To address this, uni-robot-agent has been introduced as a unified harness for competitive open-weight robot use that supports further training.
Without privileged inputs, Qwen3.8-Flash-Next achieved a RoboDojo Score of 31.39, which the authors state surpasses the published GPT-6 Astra-RoboProbe result of 28.97 and many VLA/WAM baselines. It also performs competitively on 19 real-world tasks and received 21 upvotes on Hugging Face Daily Papers.
This item is a summary. The full, checkable version is at the original source:
Read the original at Hugging Face Papers ↗