Reproduce “TaFD: Threat-Aware Frequency Decoupling for Adversarial Robustness against Heterogeneous Attacks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mengda Xie, Yiling He, Meie Fang
TaFD is an image classifier designed to handle multiple adversarial threats. It diagnoses threat domains from spectral signatures and routes image features through specialized experts, rather than relying on one shared processing path. On CIFAR-100 with ResNet-34, it achieves 31.9% Avg-RA across seven attacks, versus 25.6% for the runner-up methods; this is a benchmark result, not evidence for all attacks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.