Premise Order Matters in Reasoning with Large Language Models

TL;DR

This paper reveals that large language models' reasoning accuracy significantly depends on the order of premises, with correct ordering greatly improving performance in deductive reasoning and mathematical problem-solving tasks.

Contribution

The study systematically demonstrates the impact of premise order on LLM reasoning performance and introduces the R-GSM benchmark for further evaluation.

Findings

Premise order affects LLM reasoning accuracy by over 30%.

Aligning premise order with reasoning steps improves performance.

Introducing R-GSM benchmark for ordering effects in math problems.

Abstract

Large language models (LLMs) have accomplished remarkable reasoning performance in various domains. However, in the domain of reasoning tasks, we discover a frailty: LLMs are surprisingly brittle to the ordering of the premises, despite the fact that such ordering does not alter the underlying task. In particular, we observe that LLMs achieve the best performance when the premise order aligns with the context required in intermediate reasoning steps. For example, in deductive reasoning tasks, presenting the premises in the same order as the ground truth proof in the prompt (as opposed to random ordering) drastically increases the model's accuracy. We first examine the effect of premise ordering on deductive reasoning on a variety of LLMs, and our evaluation shows that permuting the premise order can cause a performance drop of over 30%. In addition, we release the benchmark R-GSM, based on GSM8K, to examine the ordering effect for mathematical problem-solving, and we again observe a significant drop in accuracy, relative to the original GSM8K benchmark.