Being-H0.7: A Latent World-Action Model from Egocentric Videos

2026-04-30

AI Summary

This work addresses the limitations of existing vision-language-action (VLA) models, which often exploit shortcuts due to sparse action supervision, and pixel-based future prediction approaches, which are computationally expensive and contain redundant information. The authors propose a latent-space world-action model that, for the first time, integrates future-aware reasoning into VLA policies. By leveraging learnable latent queries and a prior-posterior dual-branch architecture, the model aligns future information in latent space without explicitly generating video frames. This approach synergistically combines the predictive power of world models with the efficiency of direct policy learning, achieving state-of-the-art or competitive performance across six simulation benchmarks and multiple real-world tasks, while maintaining computational efficiency and deployability.

Abstract

Visual-Language-Action models (VLAs) have advanced generalist robot control by mapping multimodal observations and language instructions directly to actions, but sparse action supervision often encourages shortcut mappings rather than representations of dynamics, contact, and task progress. Recent world-action models introduce future prediction through video rollouts, yet pixel-space prediction is a costly and indirect substrate for control, as it may model visual details irrelevant to action generation and introduces substantial training or inference overhead. We present Being-H0.7, a latent world-action model that brings future-aware reasoning into VLA-style policies without generating future frames. Being-H0.7 inserts learnable latent queries between perception and action as a compact reasoning interface, and trains them with a future-informed dual-branch design: a deployable prior branch infers latent states from the current context, while a training-only posterior branch replaces the queries with embeddings from future observations. Jointly aligning the two branches at the latent reasoning space leads the prior branch to reason future-aware, action-useful structure from current observations alone. At inference, Being-H0.7 discards the posterior branch and performs no visual rollout. Experiments across six simulation benchmarks and diverse real-world tasks show that Being-H0.7 achieves state-of-the-art or comparable performance, combining the predictive benefits of world models with the efficiency and deployability of direct VLA policies.

Problem

• visual-language-action models
• world-action models
• future prediction
• action supervision
• latent reasoning

Innovation

• latent world-action model
• future-aware reasoning
• visual-language-action models
• dual-branch training
• egocentric video