Ambiguity-Aware In-Context Learning with Large Language Models

TL;DR

This paper proposes an ambiguity-aware demonstration selection method for in-context learning with large language models, improving performance by considering the model's existing knowledge and label ambiguity.

Contribution

It introduces a novel demonstration selection strategy that accounts for the LLM's knowledge and label ambiguity, outperforming traditional semantic similarity approaches.

Findings

Selecting demonstrations that resolve label ambiguity improves accuracy.

Including demonstrations previously misclassified by the LLM enhances performance.

Considering the model's knowledge about the task leads to better demonstration choices.

Abstract

In-context learning (ICL) i.e. showing LLMs only a few task-specific demonstrations has led to downstream gains with no task-specific fine-tuning required. However, LLMs are sensitive to the choice of prompts, and therefore a crucial research question is how to select good demonstrations for ICL. One effective strategy is leveraging semantic similarity between the ICL demonstrations and test inputs by using a text retriever, which however is sub-optimal as that does not consider the LLM's existing knowledge about that task. From prior work (Lyu et al., 2023), we already know that labels paired with the demonstrations bias the model predictions. This leads us to our hypothesis whether considering LLM's existing knowledge about the task, especially with respect to the output label space can help in a better demonstration selection strategy. Through extensive experimentation on three text classification tasks, we find that it is beneficial to not only choose semantically similar ICL demonstrations but also to choose those demonstrations that help resolve the inherent label ambiguity surrounding the test example. Interestingly, we find that including demonstrations that the LLM previously mis-classified and also fall on the test example's decision boundary, brings the most performance gain.