Skip to content

Llama-3-Patronus-Lynx-70B-Instruct

Detect RAG hallucinations with a Llama-3 model

Visit official site

What llama-3-patronus-lynx-70b-instruct is

The model is a Llama-3 based evaluator that checks answers in retrieval augmented generation for hallucinations against a source document.

What you can do with it

  • Verify that a financial report matches its sources
  • Check conclusions in medical literature against the text
  • Screen large text sets for claims that conflict with facts

Getting started

  • Open the Hugging Face model page and read the guide
  • Install Transformers and PyTorch from the examples
  • Prepare a question, document and answer, then run inference

Pros & cons

✓ What we like

  • Prompts and code examples help developers get started
  • The model is open and can be customized further

! What to watch out for

  • A document, question and answer must be prepared as input
  • Free use is stated, while commercial use may need contact

FAQ

What does it evaluate?

Whether an answer is faithful to a given document.

What is the token limit?

The material states up to 8000 tokens.

Last reviewed: 2026-11-15

More AI text detection tools

View all →

How we review