← All projects
AI
Ragtruth
A hallucination corpus for evaluating and building trustworthy RAG language models.
Stars
260
Forks
34
Language
Python
License
MIT
About
RAGTruth provides a curated dataset of hallucination annotations across diverse RAG tasks, enabling developers to benchmark and improve the faithfulness of retrieval-augmented generation systems. Accompanied by a published academic paper, it serves as a valuable evaluation resource for anyone building production RAG pipelines. The corpus covers multiple domains and task types, making it broadly applicable for hallucination detection research.