Vectara HHEM Leaderboard — New Open-Source LLM Hallucination Evaluation

AI Impact Summary

Vectara has created a new Hugging Face leaderboard powered by the HF leaderboard template and their Hughes Hallucination Evaluation Model (HHEM). This allows for dynamic updates and evaluation of LLMs like GPT-4, Gemini, and Llama 2, specifically measuring their propensity to hallucinate. The new leaderboard provides a mechanism for the community to contribute new models for evaluation and track performance metrics such as Factual Consistency Rate and Hallucination Rate.