H. Lu, “From Pretrain to Pain: Adversarial Vulnerability of Video Foundation Models Without Task Knowledge”, AAAI, vol. 40, no. 9, pp. 7548–7556, Mar. 2026.