Research map / Architectures and efficiency / Spiking networks
AdaS: Adaptive Gradient Descent for Spiking Transformers
Zijian Zhou, Honglin Cao, Ammar Belatreche, Wenjie Wei, Yimeng Shan, Yu Liang, Yu Yang, Shuai Wang, Yalan Ye, Malu Zhang, Yang Yang, Haizhou Li
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 AdaS adaptive momentum blending to AdamW | SpikeLM · GLUE (average) | average score: 76.5 -> 77.6 | Improved |
| add AdaS to AdamW on MNLI-matched | SpikeLM · GLUE MNLI-m | accuracy (%): 77.1 -> 77.0 | No clear effect |
| use adaptive balance weight instead of fixed alpha (ablation) | AdaS · CoLA | score: best fixed 40.9 -> adaptive 42.1 | Improved |
| use adaptive balance weight instead of fixed alpha (ablation) | AdaS · CIFAR10-DVS | accuracy (%): best fixed 84.2 -> adaptive 85.1 | Improved |
| fixed alpha = 0.9 instead of AdamW baseline (ablation) | AdaS · CIFAR10-DVS | accuracy (%): 83.5 -> 83.4 | Hurt |
| add AdaS to AdaX, QHAdam, Adamax and AdamW | adaptive optimizers · CoLA and CIFAR10-DVS | score / accuracy: AdamW +3.3/+1.6, AdaX +2.3/+0.5, QHAdam +1.5/+1.3, Adamax +0.9/+0.7 | Improved |
| add AdaS | SpikeLM optimization noise · CoLA, MRPC, RTE, STS-B | estimated parameter-update noise | Improved |
| add AdaS | vanilla optimizers · CIFAR10-DVS convergence | convergence speed | No clear effect |
| use SGD instead of adaptive optimizers | E-SpikeFormer · ADE20K | mIoU (%): 38.2 (AdamW) -> 4.6 (SGD) | Hurt |
| use AdaS instead of AdamW | memory and speed · CoLA, RTX 4090 | training speed (iter/s): 3.51 -> 3.15 | Hurt |
Closest papers by meaning
- Spiking Transformer: Introducing Accurate Addition-Only Spiking Self-Attention for Transformer (CVPR 2025 · similarity 0.92)
- STEP: A Unified Spiking Transformer Evaluation Platform for Fair and Reproducible Benchmarking (NeurIPS 2025 · similarity 0.91)
- Towards Efficient Spiking Transformer: a Token Sparsification Framework for Training and Inference Acceleration (ICML 2024 · similarity 0.90)
- Adaptive Calibration: A Unified Conversion Framework of Spiking Neural Networks (AAAI 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)
- Quantized Spike-driven Transformer (ICLR 2025 · similarity 0.88)
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.
