EBT-Policy: Energy Unlocks Emergent Physical Reasoning Capabilities

TL;DR

EBT-Policy introduces an energy-based model for robotic control that outperforms diffusion policies in efficiency, robustness, and emergent capabilities like zero-shot recovery, demonstrating significant advances in physical reasoning and real-world applicability.

Contribution

The paper presents EBT-Policy, a scalable energy-based architecture that enhances robustness and emergent reasoning in robotic policies, outperforming existing diffusion-based methods in efficiency and capabilities.

Findings

EBT-Policy converges within two inference steps on some tasks.

It outperforms diffusion policies in simulated and real-world tasks.

Exhibits emergent zero-shot recovery capabilities without explicit retry training.

Abstract

Implicit policies parameterized by generative models, such as Diffusion Policy, have become the standard for policy learning and Vision-Language-Action (VLA) models in robotics. However, these approaches often suffer from high computational cost, exposure bias, and unstable inference dynamics, which lead to divergence under distribution shifts. Energy-Based Models (EBMs) address these issues by learning energy landscapes end-to-end and modeling equilibrium dynamics, offering improved robustness and reduced exposure bias. Yet, policies parameterized by EBMs have historically struggled to scale effectively. Recent work on Energy-Based Transformers (EBTs) demonstrates the scalability of EBMs to high-dimensional spaces, but their potential for solving core challenges in physically embodied models remains underexplored. We introduce a new energy-based architecture, EBT-Policy, that solves core issues in robotic and real-world settings. Across simulated and real-world tasks, EBT-Policy consistently outperforms diffusion-based policies, while requiring less training and inference computation. Remarkably, on some tasks it converges within just two inference steps, a 50x reduction compared to Diffusion Policy's 100. Moreover, EBT-Policy exhibits emergent capabilities not seen in prior models, such as zero-shot recovery from failed action sequences using only behavior cloning and without explicit retry training. By leveraging its scalar energy for uncertainty-aware inference and dynamic compute allocation, EBT-Policy offers a promising path toward robust, generalizable robot behavior under distribution shifts.