Event
UniWM Unified Visual Navigation World Model Accepted by ECCV 2026
1 reports · 1 sources
Yifei Dong and other researchers propose UniWM, a memory-augmented world model that integrates first-person visual foresight and planning within a single multimodal autoregressive backbone, enabling action selection based on visually imagined outcomes and achieving alignment between prediction and control. The paper has been accepted by ECCV 2026. The authors state that experiments on four benchmarks—Go Stanford, ReCon, SCAND, and HuRoN—as well as the 1X Humanoid Dataset show that UniWM improves navigation success rates by up to 30% and achieves zero-shot generalization to the unseen TartanDrive dataset.
Report timeline
- UniWM: Towards Unified World Models for Visual Navigation via Memory-Augmented Planning and Foresight
An ECCV 2026 accepted paper proposes UniWM, a memory-augmented world model that unifies egocentric visual foresight and planning within a single multimodal autoregressive backbone, grounding action selection in visually imagined outcomes and tightly aligning prediction with control.
arXiv cs.RO ·