[AAAI 2026] A PyTorch official implementation for Boosting Adversarial Transferability via Ensemble Non-Attention

This work is the first to explore the power of ensemble non-attention in improving cross-architecture transferability. We propose a novel ensemble attack, NAMEA, which integrates ensemble non-attention and meta learning to ensure stable update direction and model diversity at once.

@article{zou2025boosting,
  title={Boosting Adversarial Transferability via Ensemble Non-Attention},
  author={Zou, Yipeng and Liu, Qin and Wu, Jie and Peng, Yu and Chen, Guo and Zhou, Hui and Ye, Guanghui},
  journal={arXiv preprint arXiv:2511.08937},
  year={2025}
}