# FastVLA: Task-conditioned and temporally adaptive visual token compression for vision-language-action models

- Source: arXiv cs.RO
- Published: 2026-10-21T04:00:00.000Z
- RoboHot: https://robohot.news/en/items/h8xkf87pontft9gv5s5tnrrot
- Original: https://arxiv.org/abs/2610.22146

## Summary

FastVLA proposes a training-free, task-aware, and temporally adaptive visual token compression framework to improve the closed-loop inference efficiency of vision-language-action (VLA) models. It scores tokens via task-to-vision attention and adaptively sets the compression rate based on the uniformity of the importance distribution, calibrated using an inverse Simpson index and an online sliding window.