Skip to content

hallucination-leaderboard

Rank LLMs by hallucination rate on short summaries

Visit official site

What hallucination-leaderboard is

The project is described as an open source leaderboard that ranks LLMs by hallucination rate on short document summaries.

What you can do with it

  • Evaluate summaries with the HHEM-2.1 model
  • Compare and rank multiple LLMs
  • Read hallucination and consistency metrics

Application scenarios

  • Compare hallucination rates to pick a reliable model
  • Evaluate a self developed LLM
  • Choose an LLM for content generation or support

Pros & cons

✓ What we like

  • CNN/Daily Mail is used for summary tests
  • LLM evaluation runs through API calls

! What to watch out for

  • Multilingual evaluation currently mainly supports English
  • Models and their API endpoints must be prepared for evaluation

FAQ

Which corpus is named?

CNN/Daily Mail is listed for document summary tests.

Which metrics are listed?

Hallucination rate, factual consistency rate and answer rate are listed.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review