arXiv：cs.CL· Jingya Wang, Dehao Zhang, Shuai Wang, Malu Zhang, Yang Yang, Haizhou Li·· 3 小时前AI 评分34
SpikingVLA：异步脉冲视觉-语言-动作模型
SpikingVLA: Asynchronous Spiking Vision-Language-Action Models
AI 导读
SpikingVLA 是一个 ANN-to-SNN 转换框架，通过提出 Dendritic Integrate-and-Fire（DIF）神经元和异步执行机制，实现低延迟的脉冲 VLA 推理。相比现有脉冲 VLA 方法，其 SR 和 SPL 分别提升 11.9% 和 12.6%，首动作延迟降低 11.2 倍。
正文
Abstract:ANN-to-SNN conversion offers a practical route toward energy-efficient spiking Vision-Language-Action (VLA) models by bypassing the substantial cost of training large-scale SNNs from scratch. However, existing methods often require many timesteps to maintain competitive performance, resulting in substantial inference latency for real-time VLA deployment. To address this challenge, we introduce SpikingVLA, an ANN-to-SNN conversion framework that enables accurate and low-latency spiking VLA inference. Specifically, we propose a Dendritic Integrate-and-Fire (DIF) neuron that alleviates channel-wise activation outliers through dendritic mixing and adaptive somatic firing, enabling accurate ANN-to-SNN conversion with fewer timesteps. Building on DIF neurons, we further introduce an asynchronous execution mechanism that overlaps temporal computation across VLA components, reducing synchronization overhead and latency. Extensive experiments demonstrate that SpikingVLA achieves competitive navigation performance with substantially improved inference efficiency. Compared with existing spiking VLA methods, SpikingVLA improves SR and SPL by 11.9\% and 12.6\%, respectively, while reducing first-action latency by 11.2$\times$. These results establish SpikingVLA as a practical framework for deploying pretrained VLA models with high-performance and low-latency spiking inference.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.09710 [cs.CL] |
| (or arXiv:2610.09710v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.09710 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jingya Wang [view email]
[v1]
Wed, 7 Oct 2026 09:05:07 UTC (12,941 KB)
来源：arXiv：cs.CL · arxiv.org