FaithBench: A Diverse Hallucination Benchmark for Summarization by Modern LLMs
This addresses the lack of diversity and recency in hallucination benchmarks for summarization, which is crucial for improving LLM reliability in applications like RAG, though it is incremental as it builds on existing evaluation frameworks.
The paper tackles the problem of evaluating hallucinations in LLM-generated summaries by introducing FaithBench, a benchmark with diverse and challenging hallucinations from 10 modern LLMs, showing that even the best detection models achieve only near 50% accuracy.
Summarization is one of the most common tasks performed by large language models (LLMs), especially in applications like Retrieval-Augmented Generation (RAG). However, existing evaluations of hallucinations in LLM-generated summaries, and evaluations of hallucination detection models both suffer from a lack of diversity and recency in the LLM and LLM families considered. This paper introduces FaithBench, a summarization hallucination benchmark comprising challenging hallucinations made by 10 modern LLMs from 8 different families, with ground truth annotations by human experts. ``Challenging'' here means summaries on which popular, state-of-the-art hallucination detection models, including GPT-4o-as-a-judge, disagreed on. Our results show GPT-4o and GPT-3.5-Turbo produce the least hallucinations. However, even the best hallucination detection models have near 50\% accuracies on FaithBench, indicating lots of room for future improvement. The repo is https://github.com/vectara/FaithBench