Salesforce/PROVE
收藏Hugging Face2025-02-03 更新2025-04-08 收录
下载链接:
https://hf-mirror.com/datasets/Salesforce/PROVE
下载链接
链接失效反馈官方服务:
资源简介:
PROVE是一个用于评估视觉语言模型在开放性问题中对视觉查询的回答的质量的基准测试数据集。它包含了10.5k个具有挑战性但基于视觉的问题-答案(QA)对,通过让大型语言模型生成QA对和验证程序来构建。数据集支持对VLM模型的帮助性和真实性进行评估。
PROVE is a benchmark test dataset for evaluating the quality of visual query responses by Vision-Language Models (VLMs) in open-ended questions. It contains 10.5k challenging yet visually grounded question-answer (QA) pairs, constructed by prompting a large language model to generate QA pairs and verification programs. The dataset supports the evaluation of helpfulness and truthfulness of VLMs.
提供机构:
Salesforce



