Reasoning Capabilities and Invariability of Large Language Models

TL;DR

This paper evaluates the reasoning abilities of large language models using a new geometric reasoning benchmark, revealing their strengths and limitations in zero-shot and chain-of-thought prompting scenarios.

Contribution

It introduces a novel benchmark dataset for simple geometric reasoning tasks and provides a comprehensive empirical analysis of LLMs' reasoning capabilities and prompt dependency.

Findings

LLMs over 70B parameters perform better in zero-shot settings

Chain-of-thought prompting can improve or impair performance depending on implementation

Significant room for improvement remains in LLM reasoning abilities

Abstract

Large Language Models (LLMs) have shown remarkable capabilities in manipulating natural language across multiple applications, but their ability to handle simple reasoning tasks is often questioned. In this work, we aim to provide a comprehensive analysis of LLMs' reasoning competence, specifically focusing on their prompt dependency. In particular, we introduce a new benchmark dataset with a series of simple reasoning questions demanding shallow logical reasoning. Aligned with cognitive psychology standards, the questions are confined to a basic domain revolving around geometric figures, ensuring that responses are independent of any pre-existing intuition about the world and rely solely on deduction. An empirical analysis involving zero-shot and few-shot prompting across 24 LLMs of different sizes reveals that, while LLMs with over 70 billion parameters perform better in the zero-shot setting, there is still a large room for improvement. An additional test with chain-of-thought prompting over 22 LLMs shows that this additional prompt can aid or damage the performance of models, depending on whether the rationale is required before or after the answer.