One paper on vision-language model robustness got accepted at NeurIPS 2023.