This dataset is designed to evaluate the truthfulness and honesty of vision-language models.

from datasets import load_dataset

dataset = load_dataset("PKU-Alignment/TruthfulVQA", split="validation")

TruthfulVQA contains the following categories of truthfulness challenges:

Visual Information Distortion Blurring / Low-Resolution Processing Concealed Features and Information… See the full description on the dataset page: https://huggingface.co/datasets/PKU-Alignment/TruthfulVQA.