LeapBot-WA: World-Anchor Action Models via Predictive Latent Alignments
2026-07-27 • Robotics
Robotics
AI summaryⓘ
The authors introduce LeapBot-WA, a new way for robots to understand and predict the world around them without needing to recreate every tiny visible detail from videos. Instead of focusing on pixel-perfect video output, their method works in a simplified hidden space that captures important physical changes. They also developed special tools to keep these hidden representations organized and easy to work with. Their approach performs very well on robot tasks, is more robust to new situations, and transfers better to real-world use than previous models.
World Action Modelslatent spaceJoint-Embedding Predictive Architecturepredictive modelingdiffusion modelsautoencodertransformerrobotic controlzero-shot robustness
Authors
Pei Liu, Nan Zheng, Lang Zhang, Daojie Peng, Yanan Zhang, Feilong Kong, Mingyue Feng, Jiachao Liu, Yaonong Wang, Qifeng Chen, Jun Ma
Abstract
World Action Models (WAMs) have emerged as a powerful paradigm for embodied intelligence, yet the prevailing reliance on pixel-level video generation creates a fundamental bottleneck. Forcing models to reconstruct task-irrelevant visual details dissipates representational capacity and renders policies vulnerable to visual distractors. In this paper, we propose LeapBot-WA, which establishes a novel Predictive-Latent paradigm for WAMs by operationalizing the Joint-Embedding Predictive Architecture (JEPA) as a World-Anchor. Departing from the traditional reliance on visual synthesis, LeapBot-WA shifts the core of world modeling to Predictive Semantic Alignment, extracting abstract physical dynamics directly within a latent foundation space. To bridge the modality gap between non-Gaussian predictive features and diffusion priors, we introduce the Isotropic Semantic Autoencoder (ISAE), which reshapes the anchor's latent space into a diffusion-friendly manifold to prevent off-manifold drift. Furthermore, we design an Asymmetric Mixture-of-Transformers (MoT) architecture. During training, an Anchor Diffusion Transformer acts as a privileged dynamics expert to guide the Action Diffusion Transformer; at inference, this heavy dynamics branch is pruned, enabling zero-overhead execution. LeapBot-WA achieves state-of-the-art performance among predictive models on LIBERO and matches top-tier generative WAMs on RoboTwin 2.0 without requiring large-scale trajectory pre-training. It further demonstrates superior zero-shot robustness to unseen environments and successful real-world transfer, establishing a highly efficient and robust latent-centric paradigm for scalable robotic control. Code: https://github.com/LeapWM/leapbot-wa.