Bridge-WA: Predicting Where and How the World Changes for Robotic Action
2026-07-02 • Robotics
Robotics
AI summaryⓘ
The authors introduce Bridge-WA, a new approach to help robots and AI better understand how their actions will change a scene. Instead of using complex models that predict every detail of the future, their method focuses on key changes that matter for actions by learning simple summaries from a bigger teacher model. Bridge-WA uses these summaries to guide decisions about where and how to act, which helps it work better especially when things look different than usual. This approach improves task success and makes the system more robust without needing expensive future image predictions during use.
vision-language-action modelsworld-action modelsfuture-change predictionaction transformermulti-source attentionspatial-temporal biasesout-of-distribution generalizationchange mapsmotion-flow mapsrobotics task success
Authors
Yongjie Bai, Hanting Wang, Mingtong Dai, Qijun Zhong, Yang Liu, Liang Lin
Abstract
General-purpose vision-language-action models benefit from large vision-language priors, but effective manipulation also requires anticipating action-relevant scene changes. Existing world-action models often rely on large generative world models or dense future rollouts, which are expensive and spend capacity on visual details weakly coupled to control. We present Bridge-WA, a lightweight world-action framework that distills a frozen future-change teacher into three compact priors: future tokens for intended outcomes, change maps for intervention support, and motion-flow maps for local transition direction. A WorldBridge conditions the action transformer on these priors through multi-source attention memories and spatial-temporal biases, while the teacher model is removed at inference. Across VLABench, RoboTwin2.0, LIBERO-Plus and real-robot evaluations, Bridge-WA improves task success, progress, and robustness, with particularly clear gains under out-of-distribution visual shifts. By focusing action generation on where and how the scene will change, Bridge-WA suppresses nuisance appearance factors such as background, lighting, and distractors, leading to better generalization without deployment-time dense future-image generation. Code and visualizations are available at: https://hcplab-sysu.github.io/BRIDGE-WA .