ETA: A New Agentic Paradigm for Embodied Tasks

2026-08-04

AI Summary

Current embodied systems rely on end-to-end mappings, which suffer from limited generalization and struggle to execute long-horizon, controllable complex tasks in unseen environments. This work proposes the Embodied Task Agent (ETA) paradigm, which extends digital agent architectures into the physical world for the first time by establishing a closed-loop system composed of a planner, tool interfaces, and the environment. Its modular design enables interchangeable planners, composable skill-based tools, auditable memory, and replayable trajectories, while providing a unified interface for both simulation and real-world deployment. The open-source implementation, OpenETA, substantially enhances robotic task generalization, interpretability, and continual learning capabilities.

Abstract

When will robots have their ChatGPT moment? Such a breakthrough requires a general-purpose robot that can handle unfamiliar tasks in unfamiliar environments, remain controllable over long interactions, and learn from experience. Today's embodied systems largely follow an end-to-end observation-to-action path. Despite rapid progress, they remain far from this goal: their generalization depends heavily on the coverage of robot training data, while long task execution remains difficult to control and inspect. To realize this goal, we introduce the Embodied Task Agent (ETA), a new paradigm for extending digital agents into the physical world, and release OpenETA as its open-source implementation. ETA centers the robot around a Planner that chooses one Tool call at a time, an Interface that controls execution, and a World that returns the result and a fresh observation. This loop allows the agent to verify outcomes, adapt its plan, and turn successful and failed interactions into reusable experience. OpenETA provides replaceable Planners, composable Tools and Skills, auditable memory, replayable trajectories, and common interfaces for simulation and real robots. For Codex, OpenETA can operate as a lightweight plugin that exposes only observe, mark_point, and move_to.

Problem

• embodied intelligence
• task generalization
• long-horizon control
• experience learning
• robot autonomy

Innovation

• Embodied Task Agent
• modular planning
• tool-augmented robotics
• replayable experience
• open-source embodied AI