Knowledge of Knowledge: Exploring Known-Unknowns Uncertainty with Large Language Models

TL;DR

This paper explores how large language models understand and articulate their uncertainty regarding known-unknown questions, introducing a new dataset and demonstrating improved performance in uncertainty detection and expression.

Contribution

It introduces a novel dataset of known-unknown questions and a framework for analyzing uncertainty, along with fine-tuning methods that enhance LLMs' ability to identify and express uncertainty.

Findings

Fine-tuned models show significant F1-score improvements in uncertainty detection.

Models better distinguish between known and unknown questions after fine-tuning.

Enhanced uncertainty articulation improves multi-agent debate performance.

Abstract

This paper investigates the capabilities of Large Language Models (LLMs) in the context of understanding their knowledge and uncertainty over questions. Specifically, we focus on addressing known-unknown questions, characterized by high uncertainty due to the absence of definitive answers. To facilitate our study, we collect a new dataset with Known-Unknown Questions (KUQ) and establish a categorization framework to clarify the origins of uncertainty in such queries. Subsequently, we examine the performance of open-source LLMs, fine-tuned using this dataset, in distinguishing between known and unknown queries within open-ended question-answering scenarios. The fine-tuned models demonstrated a significant improvement, achieving a considerable increase in F1-score relative to their pre-fine-tuning state. Through a comprehensive analysis, we reveal insights into the models' improved uncertainty articulation and their consequent efficacy in multi-agent debates. These findings help us understand how LLMs can be trained to identify and express uncertainty, improving our knowledge of how they understand and express complex or unclear information.