VAG: Dual-Stream Video-Action Generation for Embodied Data Synthesis
2026-04-10 • Robotics
RoboticsComputer Vision and Pattern Recognition
AI summaryⓘ
The authors address the challenge of creating large amounts of paired video and action data for training robot policies, which is usually hard and costly to collect. They propose VAG, a new model that generates videos and corresponding robot actions together in a synchronized way, improving alignment between what is seen and what is done. Their approach works both in simulations and real environments, producing data that helps robots learn better and perform tasks more reliably. This method offers a more efficient way to create synthetic training data for robotic learning.
robot foundation modelsteleoperation datasynthetic dataWorld Modelsaction trajectoriesflow-matchingvideo-action alignment3D poolingpolicy generalizationembodied data synthesis
Authors
Xiaolei Lang, Yang Wang, Yukun Zhou, Chaojun Ni, Kerui Li, Jiagang Zhu, Tianze Liu, Jiajun Lv, Xingxing Zuo, Yun Ye, Guan Huang, Xiaofeng Wang, Zheng Zhu
Abstract
Recent advances in robot foundation models trained on large-scale human teleoperation data have enabled robots to perform increasingly complex real-world tasks. However, scaling these systems remains difficult because collecting task-specific demonstrations is expensive and labor-intensive. Synthetic data, especially generated videos, offer a promising direction, but existing World Models (WMs) are not directly suitable for policy learning since they do not provide paired action trajectories. World-Action (WA) models partially address this by predicting actions with visual outputs, yet often lack strong video-action alignment, while two-stage pipelines that generate video first and then infer actions introduce inefficiency and error accumulation. To address these limitations, we propose VAG, a unified flow-matching-based dual-stream framework that jointly generates video and action under visual and language conditioning. By synchronizing denoising in both branches and using an adaptive 3D pooling mechanism to transfer compact global video context to the action branch, VAG improves cross-modal consistency during generation. Across both simulated and real-world settings, VAG produces aligned video-action pairs with competitive prediction quality, supports executable trajectory replay, and provides useful synthetic pretraining data that improves downstream policy generalization, indicating its potential as a practical world-action model for embodied data synthesis.