Unlocking the Power of LLM Uncertainty for Active In-Context Example Selection

TL;DR

This paper introduces Unc-TTP, a novel method for classifying LLM uncertainty using output inconsistency, which improves active in-context example selection and enhances LLM performance in real-world tasks.

Contribution

The paper proposes Unc-TTP, a new inconsistency-based uncertainty classification method that outperforms existing strategies in active example selection for LLMs.

Findings

Unc-TTP effectively classifies LLM uncertainty using output inconsistency.

Uncertainty-based example selection with Unc-TTP yields more informative in-context examples.

The method improves LLM performance in real-world tasks through better uncertainty handling.

Abstract

Large Language Models (LLMs) have shown remarkable performance across a wide range of downstream tasks. However, it is challenging for users to discern whether the responses of LLM are generated with certainty or are fabricated to meet user expectations. In this paper, we introduce Uncertainty Tripartite Testing Paradigm (Unc-TTP), a novel method for classifying LLM uncertainty by leveraging output inconsistency. Specifically, Unc-TTP performs three rounds of sampling under varying label injection interference, enumerating all possible outcomes, and uses the degree of output inconsistency as the indicator of the LLM's intrinsic uncertainty. To validate the effectiveness of this inconsistency-defined uncertainty, we draw inspiration from Active Learning, comparing the informativeness of actively selected in-context examples. Our experiments show that uncertainty examples selected via Unc-TTP are more informative than certainty examples. Furthermore, the Unc-TTP-guided uncertainty-based active example selection strategy outperforms existing methods, highlighting its effectiveness in classifying LLM uncertainty and enhancing in-context learning. This work not only underscores the potential of inconsistency-based uncertainty classification for both open- and closed-source LLMs but also presents a practical approach for leveraging uncertainty to improve LLM performance in real-world tasks.