UBench: Benchmarking Uncertainty in Large Language Models with Multiple Choice Questions

TL;DR

UBench introduces a confidence interval-based benchmark for evaluating the uncertainty of large language models across diverse tasks, enabling effective comparison without requiring internal model access or extensive retraining.

Contribution

The paper presents UBench, a novel benchmark for LLM uncertainty evaluation using confidence intervals, applicable to both open and closed-source models, and explores factors affecting uncertainty estimation.

Findings

Confidence interval methods effectively quantify uncertainty.

Open-source models perform competitively with closed-source models.

Chain-of-Thought and role-playing prompts can improve uncertainty estimates.

Abstract

Despite recent progress in systematic evaluation frameworks, benchmarking the uncertainty of large language models (LLMs) remains a highly challenging task. Existing methods for benchmarking the uncertainty of LLMs face three key challenges: the need for internal model access, additional training, or high computational costs. This is particularly unfavorable for closed-source models. To this end, we introduce UBench, a new benchmark for evaluating the uncertainty of LLMs. Unlike other benchmarks, UBench is based on confidence intervals. It encompasses 11,978 multiple-choice questions spanning knowledge, language, understanding, and reasoning capabilities. Based on this, we conduct extensive experiments. This includes comparisons with other advanced uncertainty estimation methods, the assessment of the uncertainty of 20 LLMs, and an exploration of the effects of Chain-of-Thought (CoT) prompts, role-playing (RP) prompts, and temperature on model uncertainty. Our analysis reveals several crucial insights: 1) Our confidence interval-based methods are highly effective for uncertainty quantification; 2) Regarding uncertainty, outstanding open-source models show competitive performance versus closed-source models; 3) CoT and RP prompts present potential ways to improve model reliability, while the influence of temperature changes follows no universal rule. Our implementation is available at https://github.com/Cyno2232/UBENCH.