Zetta $ζ$: An Efficient Closed-Loop Embodied Harness for Self-Evolving Physical Intelligence

2026-08-17

AI Summary

This study addresses the lack of closed-loop learning and real-time adaptability in embodied agents by proposing Zetta, a framework featuring a novel tri-timescale closed-loop self-evolution mechanism. By employing a code-level runtime evaluator and verification gating, Zetta enables action-level governance and online skill evolution while keeping the base policy frozen. Experiments demonstrate that Zetta achieves success rates of 90.8% on LIBERO-Pro and 93.6% on RoboCasa, alongside an 11.1× inference speedup. Furthermore, the framework facilitates zero-shot skill transfer and continuous optimization, significantly enhancing the reliability and adaptive capabilities of physical intelligence systems.

Abstract

Embodied agents are increasingly used to close the gap left by end-to-end policy models. Yet the agentic path has not realized closed-loop learning in physical execution: existing harnesses remain largely open-loop, following fixed skills during rollout and reflecting only after an episode completes. Such post-hoc reflection cannot govern execution as it unfolds, because physical interaction requires decisions to track rapidly changing robot-environment states at a frequency beyond today's large agentic models. We present Zetta, a closed-loop embodied harness that evolves code-based runtime critics and recovery skills online while keeping the base policy frozen. Through three timescale-separated loops, Zetta provides action-frequency governance, rollout-level critic-recovery proposal, and validation-gated skill updates. Together with Z-Infra, a rollout infrastructure decoupling agent logic from heterogeneous execution resources, Zetta achieves state-of-the-art success on LIBERO-Pro and RoboCasa under our current rollout budget, reaching 90.8% and 93.6%, with an 11.1x inference speedup; success continues to scale with self-exploration experience; learned skills transfer zero-shot, and clear robotic "Aha Moments" emerge. These results show that closed-loop harness self-evolution opens a scaling path for reliable physical intelligence.

Problem

• Closed-loop learning
• Embodied agents
• Physical intelligence
• Real-time execution
• Self-evolution

Innovation

• Closed-Loop Embodied Harness
• Self-Evolving Physical Intelligence
• Runtime Critics
• Three Timescale-Separated Loops
• Zero-Shot Transfer