Model Releases
Reducing Temporal Redundancy for Efficient Vision-Language-Action Inference
arXiv:2607.12287v1 Announce Type: new Abstract: Vision-Language-Action (VLA) models exhibit strong generalization for robotic manipulation, yet their high inference latency limits real time deployment
arXiv:2607.12287v1 Announce Type: new Abstract: Vision-Language-Action (VLA) models exhibit strong generalization for robotic manipulation, yet their high inference latency limits real time deployment. We identify two primary sources of temporal redundancy in existing VLA pipelines: repeated visual encoding of highly similar consecutive frames and multi step iterative sampling in diffusion based policies. To address this, we propose a system level acceleration strategy that reduces computation in both perception and action generation. On the perception side, we incrementally update only tokens corresponding to dynamic scene regions instead of re-encoding entire frames. On the policy side, we compress diffusion sampling into a compact 2-step schedule through efficiency oriented training while preserving action precision. Experiments on Libero, RobotWin, and Real Robot Platforms demonstrate over 2 times speedup while maintaining high performance, achieving up to 98% success rate on general manipulation benchmarks. Our codes will be released on Github.
Related
- VLA-ATTC: Adaptive Test-Time Compute for VLA Models with Relative Action Critic Model
- Embodied.cpp: A Portable Inference Runtime of Embodied AI Models on Heterogeneous Robots
- LiteVLA-H: Dual-Rate Vision-Language-Action Inference for Onboard Aerial Guidance and Semantic Perception
Source: arXiv cs.RO | 2026-07-15