Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2604.25570 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911628917735424 |
|---|---|
| author | Bai, Dewei Peng, Hongxiang Zeng, Yunyun Zhang, Ziyu Qu, Hong Zhang, Yi |
| author_facet | Bai, Dewei Peng, Hongxiang Zeng, Yunyun Zhang, Ziyu Qu, Hong Zhang, Yi |
| contents | Spiking Transformers have shown strong potential for long-range visual modeling through spike-driven self-attention. However, their quadratic token interactions remain fundamentally misaligned with the sparse and event-driven nature of spiking neural computation. To address this limitation, we propose Vision SmolMamba, an energy-efficient spiking state-space architecture that integrates spike-driven dynamics with linear-time selective recurrence. The key idea is a Spike-Guided Spatio-Temporal Token Pruner (SST-TP), which estimates token importance using both spike activation strength and first-spike latency. This mechanism progressively removes redundant tokens while preserving salient spatio-temporal information, enabling efficient scaling with token sparsity. Based on this mechanism, the proposed SmolMamba block incorporates spike events directly into bidirectional state-space recurrence, forming a spiking state-space vision backbone for efficient long-range modeling. Extensive experiments on both static and event-based benchmarks, including ImageNet-1K, CIFAR10/100, CIFAR10-DVS, and DVS128 Gesture, demonstrate that Vision SmolMamba consistently achieves superior accuracy-efficiency trade-offs. In particular, it reduces the estimated energy cost by at least 1.5x compared with prior spiking Transformer baselines and a Spiking Mamba variant while maintaining competitive or improved accuracy. These results demonstrate that combining spike-guided token sparsity with state-space modeling offers a scalable and energy-efficient paradigm for spiking vision systems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_25570 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Vision SmolMamba: Spike-Guided Token Pruning for Energy-Efficient Spiking State-Space Vision Models Bai, Dewei Peng, Hongxiang Zeng, Yunyun Zhang, Ziyu Qu, Hong Zhang, Yi Computer Vision and Pattern Recognition Spiking Transformers have shown strong potential for long-range visual modeling through spike-driven self-attention. However, their quadratic token interactions remain fundamentally misaligned with the sparse and event-driven nature of spiking neural computation. To address this limitation, we propose Vision SmolMamba, an energy-efficient spiking state-space architecture that integrates spike-driven dynamics with linear-time selective recurrence. The key idea is a Spike-Guided Spatio-Temporal Token Pruner (SST-TP), which estimates token importance using both spike activation strength and first-spike latency. This mechanism progressively removes redundant tokens while preserving salient spatio-temporal information, enabling efficient scaling with token sparsity. Based on this mechanism, the proposed SmolMamba block incorporates spike events directly into bidirectional state-space recurrence, forming a spiking state-space vision backbone for efficient long-range modeling. Extensive experiments on both static and event-based benchmarks, including ImageNet-1K, CIFAR10/100, CIFAR10-DVS, and DVS128 Gesture, demonstrate that Vision SmolMamba consistently achieves superior accuracy-efficiency trade-offs. In particular, it reduces the estimated energy cost by at least 1.5x compared with prior spiking Transformer baselines and a Spiking Mamba variant while maintaining competitive or improved accuracy. These results demonstrate that combining spike-guided token sparsity with state-space modeling offers a scalable and energy-efficient paradigm for spiking vision systems. |
| title | Vision SmolMamba: Spike-Guided Token Pruning for Energy-Efficient Spiking State-Space Vision Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2604.25570 |