Zhao, X.; Shi, J.; Li, Y.; Huang, J.; Fan, C. Good Gradients Poison Your Model: Evading Defenses in Federated Learning via Boundary-Adaptive Perturbation. AAAI 2026, 40, 13262-13270.