FacTool: Factuality Detection in Generative AI -- A Tool Augmented
Framework for Multi-Task and Multi-Domain Scenarios, I-Chun Chern+, N/A, arXiv'23 #906
The emergence of generative pre-trained models has facilitated the synthesisof high-quality text, but it has also posed challenges in identifying factualerrors in the generated text. In particular: (1) A wider range of tasks nowface an increasing risk of containing factual errors when handled by generativemodels. (2) Generated texts tend to be lengthy and lack a clearly definedgranularity for individual facts. (3) There is a scarcity of explicit evidenceavailable during the process of fact checking. With the above challenges inmind, in this paper, we propose FacTool, a task and domain agnostic frameworkfor detecting factual errors of texts generated by large language models (e.g.,ChatGPT). Experiments on four different tasks (knowledge-based QA, codegeneration, mathematical reasoning, and scientific literature review) show theefficacy of the proposed method. We release the code of FacTool associated withChatGPT plugin interface at https://github.com/GAIR-NLP/factool .
URL
Affiliations
Abstract
Translation (by gpt-3.5-turbo)
Summary (by gpt-3.5-turbo)