KR Labs develops transparent and trustworthy AI solutions for natural language processing (NLP) tasks. KR Labs was founded by NLP researchers Ádám Kovács and Gábor Recski. Our open-source technology is based on a novel combination of statistical and rule-based methods developed at TUW-NLP and HLT-Budapest. Between 2023 and 2024, KR Labs was incubated in the TU Wien Innovation Incubation Center (i²c).
LettuceDetect is a hallucination detection framework designed for Retrieval-Augmented Generation (RAG) systems. Built on ModernBERT, it identifies unsupported segments in LLM-generated answers in real time. It overcomes the limitations of short context windows and the high computational costs associated with large LLM-based detectors.
Learn More on GitHub • Download on PyPI • View Models on Hugging Face
Traditional RAG systems retrieve relevant documents and then allow an LLM to freely generate responses based on that context. This can lead to hallucinations where the model invents facts not present in the source material. VerbatimRAG is our RAG framework that prevents hallucination by ensuring all generated content is explicitly derived from source documents.