arXiv cs.AI· Shenglan Li, Zhendong Mi, Hengyi Zhu, Jingwu Luo, Chun Kit Chan, Geng Yuan, Yanzhi Wang, Pu Zhao, Shaoyi Huang·· 6 小时前AI 评分37
IG-VLA：通过内化时空想象实现高效 VLA 推理
Imagine the Future, Internalize the Gist: Efficient VLA Reasoning via Internalized Spatiotemporal Imagination
AI 导读
研究者提出 IG-VLA 框架，让 VLA 模型在视觉表示空间中想象未来场景演化，并将其内化为紧凑的 Scene Gist Token，从而在推理时免去显式未来想象。
正文 · 原文
Abstract:Vision-language-action (VLA) models increasingly incorporate intermediate reasoning to improve robotic manipulation, yet existing approaches primarily reason about observed states without explicitly anticipating future scene evolution. Extending such reasoning to explicit future rollouts at every inference step, however, introduces substantial computational overhead. We propose IG-VLA, a VLA reasoning framework that enables models to imagine the future and internalize the gist. Our Latent Spatiotemporal Reasoning learns to imagine task-relevant future scene evolution directly in visual representation space, guiding action prediction without costly pixel-level video generation. To further reduce inference overhead, we introduce Scene Gist Memory, which internalizes reasoning-derived scene-behavior associations into a compact Scene Gist Token, preserving the benefits of future reasoning while bypassing explicit future imagination at inference. Extensive experiments on LIBERO, LIBERO-Plus, and VLABench demonstrate the effectiveness and efficiency of IG-VLA. On the LIBERO-Plus Language suite, both the reasoning and gist policies outperform the strongest baseline by nearly 6% in success rate. The gist policy also achieves up to 6.38x speedup over baselines, reducing inference latency from 1081ms to 169.5ms per action chunk on a single NVIDIA A6000 GPU. These results demonstrate that future spatiotemporal reasoning can be effectively internalized for efficient VLA deployment.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Robotics (cs.RO) |
| Cite as: | arXiv:2610.02626 [cs.CV] |
| (or arXiv:2610.02626v4 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.02626 arXiv-issued DOI via DataCite |
Submission history
From: Zhendong Mi [view email]
[v1]
Fri, 2 Oct 2026 00:31:51 UTC (3,217 KB)
[v2]
Thu, 8 Oct 2026 13:20:13 UTC (3,217 KB)
[v3]
Tue, 13 Oct 2026 20:07:42 UTC (3,215 KB)
[v4]
Sun, 25 Oct 2026 14:35:44 UTC (3,215 KB)
来源：arXiv cs.AI · arxiv.org