Adversary-Aware DPO: Enhancing Safety Alignment in Vision Language Models via Adversarial Training.

Fenghua Weng, Jian Lou, Jun Feng, Minlie Huang, Wenjie Wang

Browse the full EMNLP paper archive.