W. Cai, “When Safe Unimodal Inputs Collide: Optimizing Reasoning Chains for Cross-Modal Safety in Multimodal Large Language Models”, AAAI, vol. 40, no. 42, pp. 35322–35330, Mar. 2026.