Research map / Architectures and efficiency / Spiking networks
SMM Transformer: Leveraging Spiking Neural Networks for Multimodal Tasks
Xiubo Liang, Jinxing Han, Yuke Li, Haoqi Zhu, Yu Zhao, Hongzhi Wang
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 |
|---|---|---|---|
| add PLMP neuron (instead of standard LIF with STBP) (ablation) | SMM Transformer · NLVR2 dev | accuracy: 76.74 -> 78.36 | Improved |
| add SMSA (instead of SSA baseline) (ablation) | SMM Transformer · NLVR2 dev | accuracy: 77.45 -> 78.36 | Improved |
| add SMoE modality-aware experts (ablation) | SMM Transformer · NLVR2 dev | accuracy: 77.55 -> 78.36 | Improved |
| add MLM pre-training task to ITC (ablation) | SMM Transformer · NLVR2 dev | accuracy: 54.74 -> 72.53 | Improved |
| add ITM pre-training task to ITC (ablation) | SMM Transformer · NLVR2 dev | accuracy: 54.74 -> 73.63 | Improved |
| make LIF parameters learnable (PLMP K=1) (ablation) | static LIF · MSCOCO captioning | CIDEr: 125.26 -> 125.69 | Improved |
| increase branches from K=1 to K=3 (ablation) | PLMP · MSCOCO captioning | CIDEr: 125.69 -> 127.72 | Improved |
| increase branches from K=3 to K=4 (ablation) | PLMP · MSCOCO captioning | CIDEr: 127.72 -> 127.63 | No clear effect |
| multi-stage instead of shared alpha initialization (ablation) | PLMP K=3 · MSCOCO captioning | CIDEr: 127.72 -> 126.81 | Mixed |
| add self-compensation branch to SMSA (ablation) | SMSA · NLVR2 | accuracy: 76.97 -> 77.53 | Improved |
| use shared SMSA (with SMoE) instead of separate SMSA for image and text (ablation) | SMM Transformer · NLVR2 dev | accuracy: 77.53 -> 79.17 | Improved |
| replace dense softmax self-attention with SMSA (ablation) | SMM Transformer · NLVR2 | accuracy: 77.80 -> 77.53 | Hurt |
Closest papers by meaning
- Spiking Transformer: Introducing Accurate Addition-Only Spiking Self-Attention for Transformer (CVPR 2025 · similarity 0.89)
- Spike-driven Transformer V2: Meta Spiking Neural Network Architecture Inspiring the Design of Next-generation Neuromorphic Chips (ICLR 2024 · similarity 0.88)
- SpikedAttention: Training-Free and Fully Spike-Driven Transformer-to-SNN Conversion with Winner-Oriented Spike Shift for Softmax Operation (NeurIPS 2024 · similarity 0.88)
- Quantized Spike-driven Transformer (ICLR 2025 · similarity 0.87)
- Towards Efficient Spiking Transformer: a Token Sparsification Framework for Training and Inference Acceleration (ICML 2024 · similarity 0.87)
- Spiking Vision Transformer with Saccadic Attention (ICLR 2025 · similarity 0.87)
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.
