README.md

metadata

license: mit
task_categories:
  - text-generation
language:
  - en
tags:
  - safety
  - risk
size_categories:
  - n<1K

This is the dataset in the paper "Towards Evaluating Proactive Risk Awareness of Multimodal Language Models", and it is divided into three files:

• multi_modal_eval_hugging.json
  Contains 128 image-based samples.

• combine_images.zip
  Includes 128 image files corresponding to the samples.

• text_eval_hugging.json
  Contains 288 text-based samples.