Xi Ye, Yiwen Liu, Lina Wang +4
Text-to-image (T2I) models have raised increasing safety concerns due to their capacity to generate NSFW and other banned objects. To mitigate these risks, safety filters and conce…