RobotUse: Allocating Computation, Context, and Decisions
RobotUse lets robot agents revise physical actions, reaching 45% task success on RoboLab.
RobotUse is a robot-agent harness that lets agents visually choose targets and poses while a backend handles geometry, motion planning, and control. Subagents retain subgoal detail and return only what later decisions need, and a persistent playbook is updated from execution. On RoboLab it reaches 45% task success, 6.7 percentage points above CaP-X, and can learn from imperfect real-world feedback and transfer that experience to later tasks.
- Agents select targets and poses; backend plans and controls motion.
- Persistent playbook updates from execution across attempts.
- 45% RoboLab success, 6.7 points above CaP-X.
- Learns from imperfect real-world feedback and transfers to new tasks.
From the original article
1 min read at huggingface.co
Robot agents must connect their intended actions to observed outcomes while retaining the context needed to revise their choices over repeated attempts. Existing interfaces often leave these choices inside predefined tools or require agents to manage detailed execution code and its growing history.
By Junhoo Lee, Injun Baek, Seungyeon Kim · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.