The dataset was created as part of the paper: How Much Do LLMs Hallucinate across Languages? On Multilingual Estimation of LLM Hallucination in the Wild
Below is the figure summarizing the multilingual hallucination detection dataset creation (and multilingual hallucination evaluation dataset):
The dataset is a multilingual extension of FAVA. The dataset is created by sourcing 150 prompts from
FAVA… See the full description on the dataset page:
https://huggingface.co/datasets/WueNLP/mHallucination_Detection.
