Abstract:Spiking Neural Networks (SNNs) offer notable advantages in biological plausibility and energy efficiency, making them promising candidates for building low-power Transformers. However, existing Spiking Transformers largely adhere to a passive reactive paradigm, which struggles to focus on task-relevant information and incurs substantial computational overhead when processing redundant visual data. To overcome this fundamental yet underexplored limitation, we propose SAFformer, a novel Spiking Transformer architecture based on an active predictive filtering paradigm. Inspired by the brain's predictive coding mechanism, SAFformer actively suppresses predictable signals and focuses on salient visual features. Extensive experiments show that SAFformer establishes new state-of-the-art performance on CIFAR-10/100 and CIFAR10-DVS. Remarkably, on ImageNet-1K, it achieves 80.50% Top-1 accuracy with only 26.58M parameters and an energy consumption of 5.88 mJ, demonstrating an exceptional balance between accuracy and efficiency.
| Comments: | 12pages,7pages,ijcai2026 |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| MSC classes: | 68T07 |
| ACM classes: | I.2.10 |
| Cite as: | arXiv:2605.08270 [cs.CV] |
| (or arXiv:2605.08270v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2605.08270 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zequan Xie [view email]
[v1]
Fri, 8 May 2026 03:17:52 UTC (3,405 KB)