Zhang, X. et al. 2026. ShadeEdit: A Utility-Preserving and Defense-Evasive Knowledge Manipulation Attack in Federated LLMs. Proceedings of the AAAI Conference on Artificial Intelligence. 40, 41 (Mar. 2026), 34845–34853. DOI:https://doi.org/10.1609/aaai.v40i41.40787.