AI Daybook
AI Daybook
Today's issue
PaperPhysical AI

WoVR+: world model as reliable simulator for VLA policy optimization with RL

Original titleWoVR+: World Model as Reliable Simulator for Versatile VLA Policy Optimization with RL

Outlet
arXiv
Published
Importance
2/5
Trending
3

Summary

WoVR+ extends a hallucination-aware world-model RL framework by redesigning the learned environment and imagined interaction protocol for Vision-Language-Action models. It improves success over supervised fine-tuning by 21.09% and 16.7% on RoboTwin and six real-robot tasks, and raises long-horizon phone-packaging success from 30% to 75%.

Also reported by 1 outlet

Original excerpt

Action-conditioned world models can reduce costly real-robot interaction in reinforcement learning for Vision-Language-Action models (VLA-RL), but hallucinations can mislead policy optimization. Extending them to richer VLA interfaces and long-horizon tasks raises challenges in spatiotemporal and cross-modal consistency, exploration-reliability balance in RL, and compounding rollout errors. We present WoVR+, extending WoVR's hallucination-aware WM-RL framework by redesigning the learned environment and imagined interaction protocol: A VLA-interface-aligned world model provides multi-view prediction and proprioceptive updates while mitigating hallucinations. Chunk-wise noise preserves flow-based policy exploration while limiting prediction-degrading action jitter. For long-horizon tasks, stage-local imagined training combines stage-entry resets, stage-specific rewards, local horizons, and controllable sampling to shorten generated prefixes and balance training across stages. WoVR+ improves success over SFT by 21.09% and 16.7% on RoboTwin and 6 real-robot tasks. Long-horizon phone-packaging success rises from 30% to 75%. Videos are available at https://youmo445.github.io/WoVR_Plus.github.io/.

#vla#world-model#robotics#reinforcement-learning
Read the original
See all →