AI Paper Atlas

Research map / Architectures and efficiency / Spiking networks

Training-Free ANN-to-SNN Conversion for High-Performance Spiking Transformers

Jingya Wang, Xin Deng, Wenjie Wei, Dehao Zhang, Shuai Wang, Qian Sun, Jieyuan Zhang, Hanwen Liu, Ning Xie, Malu Zhang

AAAI 2026 · Architectures and efficiency › Spiking networks · approach: Training deep spiking networks · 8 citations (OpenAlex)

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 changedWhereMeasuredEffect
replace Few-Spikes neurons with MBE neurons for GELU conversion (ablation)Transformer ANN-to-SNN conversion · GELU conversion accuracyaccuracy: +13.20%Improved
replace FS neurons with MBE neurons for FP multiplication, Softmax and LayerNorm (ablation)Transformer ANN-to-SNN conversion · Transformer componentsaccuracyImproved
increase number of bases N in MBE neuron (ablation)MBE neuron replacing GELU · ViT GELU replacementaccuracyImproved
choose N=4 rather than larger N (ablation)MBE neuron · accuracy vs energy trade-offaccuracy-energy marginal rateNo clear effect
add exponential decay to multi-basis encoding (ablation)MBE neuron · function approximation, N=1,2,4,6,8log MSE: MSE reduced by 1-2 orders of magnitudeImproved
use binary initialization instead of random for FS neurons (ablation)FS neuron approximating ReLU · approximation MSEMSE: 1.5 x 10^-3 -> 9.4 x 10^-5Improved
increase timesteps from 12 to 16 (ablation)MBE conversion · ImageNet ViT-B/16top-1 accuracy (%): 82.79 -> 83.00Improved
reduce timesteps from 16 to 10 (ablation)MBE conversion · ImageNet ViT-M/16top-1 accuracy (%): 85.31 -> 21.99Hurt
reduce timesteps from 16 to 8 (ablation)MBE conversion · WikiText-103perplexity: 23.41 -> 41072Hurt
use MBE conversion (training-free, T=16) instead of SpikeZIP-TF (T=64/128) on RoBERTaRoBERTa NLU conversion · SST-2 (RoBERTa-Large)accuracy (%): 93.79 -> 95.98Improved
use MBE conversion instead of directly trained SpikeGPTGPT-2 NLG · WikiText-103perplexity: 39.75 -> 23.41Improved
use MBE conversion instead of directly trained SpikeGPTGPT-2 NLG · WikiText-2perplexity: 18.01 -> 22.69Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy