WikiBio GPT-3 Hallucination Dataset

Introduced 2023-03-15

The WikiBio GPT-3 Hallucination Dataset is a benchmark dataset used for hallucination detection. It is based on Wikipedia biographies (WikiBio) and is specifically designed to evaluate the factuality of text generated by large language models like GPT-3¹². Here are some key details about this dataset:

  • Dataset Source: Wikipedia biographies (WikiBio)
  • Task: Text classification
  • Language: English
  • Size Categories: Less than 1,000 samples
  • License: Creative Commons Attribution-ShareAlike 3.0 (cc-by-sa-3.0)

(1) potsawee/wiki_bio_gpt3_hallucination · Datasets at Hugging Face. https://huggingface.co/datasets/potsawee/wiki_bio_gpt3_hallucination. (2) [2303.08896] SelfCheckGPT: Zero-Resource Black-Box Hallucination .... https://arxiv.org/abs/2303.08896. (3) AIトラストと、対話型生成AIにおける富士通のAIトラスト技術 : 富士通. https://www.fujitsu.com/jp/about/research/article/202312-ai-trust-technologies.html. (4) README.md · potsawee/wiki_bio_gpt3_hallucination at main - Hugging Face. https://huggingface.co/datasets/potsawee/wiki_bio_gpt3_hallucination/blob/main/README.md. (5) undefined. https://github.com/potsawee/selfcheckgpt.