cambridgeltl/DARE
收藏Hugging Face2025-02-11 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/cambridgeltl/DARE
下载链接
链接失效反馈官方服务：
资源简介：
DARE（多样化的视觉问答与鲁棒性评估）是一个精心创建和策划的多选视觉问答基准。它评估VLM在五个不同类别上的性能，并包括基于提示、答案选项子集、输出格式和正确答案数量的四种鲁棒性导向评估。
DARE (Diverse Visual Question Answering with Robustness Evaluation) is a carefully created and curated multiple-choice VQA benchmark. It evaluates VLM performance on five diverse categories and includes four robustness-oriented evaluations based on the variations of prompts, subsets of answer options, output format, and the number of correct answers.
提供机构：
cambridgeltl
