Benchmarking Hallucination in Large Language Models based on Unanswerable Math Word Problem

TL;DR

This paper introduces a new benchmark and dataset for evaluating hallucination in large language models using unanswerable math word problems, demonstrating that specific training methods reduce hallucination.

Contribution

It presents the UMWP dataset and a novel evaluation methodology to assess LLM hallucination in math QA, highlighting the impact of in-context learning and RLHF.

Findings

In-context learning reduces hallucination.

RLHF training improves model reliability.

UMWP is effective for hallucination assessment.

Abstract

Large language models (LLMs) are highly effective in various natural language processing (NLP) tasks. However, they are susceptible to producing unreliable conjectures in ambiguous contexts called hallucination. This paper presents a new method for evaluating LLM hallucination in Question Answering (QA) based on the unanswerable math word problem (MWP). To support this approach, we innovatively develop a dataset called Unanswerable Math Word Problem (UMWP) which comprises 5200 questions across five categories. We developed an evaluation methodology combining text similarity and mathematical expression detection to determine whether LLM considers the question unanswerable. The results of extensive experiments conducted on 31 LLMs, including GPT-3, InstructGPT, LLaMA, and Claude, demonstrate that in-context learning and reinforcement learning with human feedback (RLHF) training significantly enhance the model's ability to avoid hallucination. We show that utilizing MWP is a reliable and effective approach to assess hallucination. Our code and data are available at https://github.com/Yuki-Asuuna/UMWP.