RoboQuest: Generalist Physical Agents that Search, Inspect and Test
The benchmark’s best frontier multimodal agent solved only 23% of RoboQuest episodes.
RoboQuest tests whether embodied agents can gather missing task evidence through physical interaction before acting. Its ten mobile manipulation tasks cover search, manipulation-based inspection, and interactive testing. The paper says agents often have the low-level skills when hidden information is supplied, but fail by deciding too early, not repairing disturbances, or struggling to learn from trial and error. HF Daily Papers' note
RoboQuest tests whether embodied agents can gather missing task evidence through physical interaction before acting. Its ten mobile manipulation tasks cover search, manipulation-based inspection, and interactive testing. The paper says agents often have the low-level skills when hidden information is supplied, but fail by deciding too early, not repairing disturbances, or struggling to learn from trial and error. HF Daily Papers' note
score 5