KnowledgePrompts: Exploring the Abilities of Large Language Models to Solve Proportional Analogies via Knowledge-Enhanced Prompting

TL;DR

This paper introduces a new dataset and evaluates large language models on proportional analogy tasks, revealing that targeted knowledge prompts significantly improve model performance, yet accuracy remains limited.

Contribution

The work presents a novel 15K MCQA dataset for proportional analogies and demonstrates the effectiveness of knowledge-enhanced prompting strategies for LLMs.

Findings

Best model achieves 55% accuracy on analogy tasks

Targeted knowledge prompts outperform other prompt types

Proportional analogy solving remains challenging for current LLMs

Abstract

Making analogies is fundamental to cognition. Proportional analogies, which consist of four terms, are often used to assess linguistic and cognitive abilities. For instance, completing analogies like "Oxygen is to Gas as <blank> is to <blank>" requires identifying the semantic relationship (e.g., "type of") between the first pair of terms ("Oxygen" and "Gas") and finding a second pair that shares the same relationship (e.g., "Aluminum" and "Metal"). In this work, we introduce a 15K Multiple-Choice Question Answering (MCQA) dataset for proportional analogy completion and evaluate the performance of contemporary Large Language Models (LLMs) in various knowledge-enhanced prompt settings. Specifically, we augment prompts with three types of knowledge: exemplar, structured, and targeted. Our results show that despite extensive training data, solving proportional analogies remains challenging for current LLMs, with the best model achieving an accuracy of 55%. Notably, we find that providing targeted knowledge can better assist models in completing proportional analogies compared to providing exemplars or collections of structured knowledge. Our code and data are available at: https://github.com/Thiliniiw/KnowledgePrompts/