vectara/hallucination-leaderboard
A public leaderboard that evaluates and ranks LLMs based on hallucination rates when summarizing documents.
Not currently ranked — collecting fresh signals.
star history
The repository provides a standardized evaluation framework for measuring how often different LLMs introduce hallucinations during document summarization. It uses Vectara’s Hallucination Evaluation Model (HHEM) to score models on metrics including hallucination rate, factual consistency rate, and answer rate. Results are regularly updated and presented as an interactive leaderboard on Hugging Face with tabular data showing performance across dozens of LLMs.
Frequently asked
- What is vectara/hallucination-leaderboard?
- A public leaderboard that evaluates and ranks LLMs based on hallucination rates when summarizing documents.
- Is hallucination-leaderboard open source?
- Yes — vectara/hallucination-leaderboard is open source, released under the Apache-2.0 license.
- What language is hallucination-leaderboard written in?
- vectara/hallucination-leaderboard is primarily written in Python.
- How popular is hallucination-leaderboard?
- vectara/hallucination-leaderboard has 3.3k stars on GitHub.
- Where can I find hallucination-leaderboard?
- vectara/hallucination-leaderboard is on GitHub at https://github.com/vectara/hallucination-leaderboard.