---
license: apache-2.0
---

This dataset contains the response of VLMs (InstructBlip, ShareGPT4V, LLaVA and CogVLM) to hateful memes and the annotation to these responses. For more information, please refer to paper "From Meme to Threat: On the Hateful Meme Understanding and Induced Hateful Content Generation in Open-Source Vision Language Models."