DVG-WM: Disentangled Video Generation Enables Efficient Embodied World Model for Robotic Manipulation

2026-06-30

AI Summary

This work addresses the challenge in existing video-based embodied world models, where the tight coupling of dynamics modeling and high-resolution visual synthesis hinders both inference efficiency and fine-grained detail fidelity, thereby limiting their applicability to dexterous robotic manipulation. To overcome this, we propose the first two-stage framework that explicitly decouples dynamics reasoning from high-fidelity video generation: it first produces a sequence of intermediate visual states conditioned on initial observations and language instructions to preview physical interactions, then refines this sequence into photorealistic video via a cascaded mechanism. By introducing flow-matching–driven direct latent mapping and a contact-detail regeneration module, our approach significantly boosts computational efficiency while preserving visual quality. Experiments demonstrate up to a 3.97× speedup in inference on both the LIBERO benchmark and real robotic platforms, alongside markedly improved video fidelity, effectively enabling planning for contact-intensive manipulation tasks.

Abstract

Video-based embodied world models provide an appealing substrate for robotic manipulation by predicting future states, yet current approaches remain limited by a fundamental entanglement: accurately modeling dynamics typically requires low-level temporal reasoning, while producing high-resolution frames demands expansive visual synthesis according to high-level semantics. This entanglement results in slow inference speed for iterative planning or too coarse predictions to retain contact-rich details. To solve this dilemma, we present Disentangled Video Generation World Model (DVG-WM), an efficient framework that explicitly decomposes world modeling into dynamics learning and visual synthesis. Conditioned on an initial observation and a language instruction, our model first generates a plausible sequence of intermediate visual states to preview the physical interaction and refines them to obtain high-fidelity videos. Furthermore, an efficient cascading mechanism is proposed, where DVG-WM uses flow matching to directly map the dynamics to video latents, and introduces a latent degradation mechanism to regenerate contact-rich details. Experiments on LIBERO and real-world platforms demonstrate improved video quality with up to 3.97 times acceleration, validating that disentangled video generation can be an efficient embodied world model for robotic manipulation.

Problem

• embodied world model
• video generation
• robotic manipulation
• dynamics modeling
• visual synthesis

Innovation

• Disentangled Video Generation
• World Model
• Flow Matching
• Latent Degradation
• Robotic Manipulation