Research map / Architectures and efficiency / Spiking networks
Temporal Interaction in Spiking Transformers with Multi-Delay Mixer
Kexin Shi, Hanwen Liu, Zeyang Song, Yang Liu, Jieyuan Zhang, Shuai Wang, Jibin Wu, Malu Zhang, Yang Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| learn delays instead of fixed random delays (ablation) | MD-Mixer on SDT-V1 · s-CIFAR10 | accuracy (%): 84.39 -> 86.65 | Improved |
| learn delays instead of fixed random delays (ablation) | MD-Mixer on SDT-V1 · CIFAR10-DVS | accuracy (%): 80.52 -> 82.44 | Improved |
| add fixed random delays only (ablation) | SDT-V1 · CIFAR10-DVS | accuracy (%): 80.00 -> 80.52 | No clear effect |
| increase number of delay branches from 1 to 3 (ablation) | MD-Mixer on SDT-V1 · s-CIFAR10 | accuracy (%): 84.3 -> 86.7 | Improved |
| increase delay branches from 3 to 6 (ablation) | MD-Mixer on SDT-V1 · s-CIFAR10 | accuracy (%): 86.7 -> 84.2 | Hurt |
| increase delay branches from 1 to 4 (ablation) | MD-Mixer on SDT-V1 · CIFAR10-DVS | accuracy (%): 81.1 -> 82.4 | Improved |
| increase timesteps from 10 to 32 (ablation) | MD-Mixer · CIFAR10-DVS | accuracy (%): 81.7 -> 83.5 | Improved |
| increase timesteps from 10 to 32 (ablation) | MD-Mixer · N-Caltech101 | accuracy (%): 84.1 -> 85.8 | Improved |
| add MD-Mixer to Spike-driven Transformer V1-8-768 | Spike-driven Transformer V1 · ImageNet-1k | top-1 accuracy (%): 76.32 -> 78.23 | Improved |
| add MD-Mixer to Spike-driven Transformer V2-8-512 | Spike-driven Transformer V2 · ImageNet-1k | top-1 accuracy (%): 79.49 -> 80.02 | Improved |
| add MD-Mixer to QKFormer | QKFormer · s-CIFAR10 | accuracy (%): 79.48 -> 85.09 | Improved |
| add MD-Mixer to SDT-V1 versus STAtten | SDT-V1 · ImageNet-1k | top-1 accuracy (%): 78.11 -> 78.23 | No clear effect |
Closest papers by meaning
- Spiking Token Mixer: An event-driven friendly Former structure for spiking neural networks (NeurIPS 2024 · similarity 0.88)
- Learning Delays in Spiking Neural Networks using Dilated Convolutions with Learnable Spacings (ICLR 2024 · similarity 0.87)
- FiTS: Interpretable Spiking Neurons via Frequency Selectivity and Temporal Shaping (NeurIPS 2026 · similarity 0.87)
- Spiking Transformer with Spatial-Temporal Attention (CVPR 2025 · similarity 0.87)
- Spiking Transformer: Introducing Accurate Addition-Only Spiking Self-Attention for Transformer (CVPR 2025 · similarity 0.86)
- Temporal Flexibility in Spiking Neural Networks: Towards Generalization Across Time Steps and Deployment Friendliness (ICLR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
