GAIR-NLP/factool
A tool-augmented framework for detecting factual errors and hallucinations in text generated by large language models.
Not currently ranked — collecting fresh signals.
star history
FacTool evaluates the factuality of LLM outputs across four task types: knowledge-based QA, code generation, mathematical reasoning, and scientific literature review. It uses external tool calls to verify claims, queries, code execution, and citations against ground truth. The framework also incorporates Halu-J, an open-source model for hallucination detection, and provides benchmarks including ChineseFactEval for evaluating LLMs in Chinese.
Frequently asked
- What is GAIR-NLP/factool?
- A tool-augmented framework for detecting factual errors and hallucinations in text generated by large language models.
- Is factool open source?
- Yes — GAIR-NLP/factool is open source, released under the Apache-2.0 license.
- What language is factool written in?
- GAIR-NLP/factool is primarily written in Python.
- How popular is factool?
- GAIR-NLP/factool has 933 stars on GitHub.
- Where can I find factool?
- GAIR-NLP/factool is on GitHub at https://github.com/GAIR-NLP/factool.