mSCoRe: a $M$ultilingual and Scalable Benchmark for $S$kill-based $Co$mmonsense $Re$asoning

TL;DR

mSCoRe is a comprehensive multilingual benchmark designed to evaluate and analyze the reasoning skills of large language models in complex, culturally diverse commonsense reasoning tasks, highlighting current limitations.

Contribution

This paper introduces mSCoRe, a novel scalable benchmark with a reasoning skill taxonomy, data synthesis pipeline, and complexity framework for multilingual commonsense reasoning evaluation.

Findings

Current LLMs find mSCoRe significantly challenging, especially at higher complexity levels.

Models show limitations in handling nuanced multilingual and cultural commonsense reasoning.

Detailed analysis suggests directions for future improvements in reasoning capabilities.

Abstract

Recent advancements in reasoning-reinforced Large Language Models (LLMs) have shown remarkable capabilities in complex reasoning tasks. However, the mechanism underlying their utilization of different human reasoning skills remains poorly investigated, especially for multilingual commonsense reasoning that involves everyday knowledge across different languages and cultures. To address this gap, we propose a \textbf{M}ultilingual and Scalable Benchmark for \textbf{S}kill-based \textbf{Co}mmonsense \textbf{Re}asoning (\textbf{mSCoRe}). Our benchmark incorporates three key components that are designed to systematically evaluate LLM's reasoning capabilities, including: (1) a novel taxonomy of reasoning skills that enables fine-grained analysis of models' reasoning processes, (2) a robust data synthesis pipeline tailored specifically for commonsense reasoning evaluation, and (3) a complexity scaling framework allowing task difficulty to scale dynamically alongside future improvements in LLM abilities. Extensive experiments on eights state-of-the-art LLMs of varying sizes and training approaches demonstrate that \textbf{mSCoRe} remains significantly challenging for current models, particularly at higher complexity levels. Our results reveal the limitations of such reasoning-reinforced models when confronted with nuanced multilingual general and cultural commonsense. We further provide detailed analysis on the models' reasoning processes, suggesting future directions for improving multilingual commonsense reasoning capabilities.