Frontier agent controls robot better with combined policy harness
Profiling a Frontier Agent with EBench: From Direct Control to Harness with Policy
RoboticsArtificial Intelligence
Summary
Robots can be controlled by language-based AI without specific robot training, but it’s unclear how well this works compared to other trained robot control methods. The authors tested a new AI agent called GPT-6-Astra on a set of robot tasks and found that it performs well on simple tasks when controlling directly but struggles on complex ones. However, when combined with specialized trained policies, it performs better overall and uses fewer computing resources. The study helps understand the strengths and limits of advanced language agents for robot control and suggests ways to improve their collaboration with trained policies.
What this means in practice
- •For robotics engineers: Improve robot control systems by integrating language-based agents with specialized policies to balance flexibility and precision in manipulation tasks.
- •For automation integrators: Develop hybrid robot control frameworks that reduce computational resources while maintaining high task success by switching between direct agent control and trained policies.
Authors
Yifei Yao, Yude Zou, Guanbo Jiao, Xing Gao, Ning Gao, Yukai Wang, Jiantong Chen, Zihou Zhu, Xinyu Li, Ningyu Yan, Shuai Wang, Jianming Ma, Haoyu Guo, Weinan Zhang, Chunhua Shen
Abstract
Frontier multimodal large language models (MLLMs) can control robots without robot-specific training, but how their capabilities compare and interact with post-trained vision-language-action (VLA) and world action model (WAM) policies across manipulation demands remains to be thoroughly evaluated. In this report, we evaluate GPT-6-Astra on EBench as the core of an agent that controls a robot through an agent harness under the same visual and proprioceptive observation protocol as post-trained policies. We study two agent configurations: direct control, in which the agent generates waypoint commands directly, and harness with policy, in which it can also invoke a post-trained VLA or WAM policy for action execution. Under direct control, the agent achieves competitive overall performance relative to post-trained policies. It leads on short-horizon tasks and exhibits the least performance variation across generalization dimensions, whereas it underperforms post-trained policies on long-horizon tasks, tabletop manipulation, and high-precision contact-rich tasks. In the harness with policy, invoking a post-trained VLA or WAM combines the complementary strengths of both components, raising overall task success to the level of the strongest compared post-trained policies while substantially reducing token cost. However, performance remains below the per-task upper bound of the standalone components due to suboptimal policy invocation, fixed-instruction execution without subgoals, and grasp disruption when the agent resumes control. Overall, our study characterizes the capability boundary of a frontier MLLM agent and suggests directions for improving future agent harness design. Project page: https://internrobotics.github.io/EBench-Astra-Tech-Report/