Year: 2026
Type: conference-paper
Abstract: Chain-of-Thought (CoT) is commonly used to improve reasoning performance in large language models.We investigate its impact in multilingual contexts by systematically constraining reasoning steps across languages with varying resource levels.This study evaluates two models on two benchmarks with seven languages, comparing constrained CoT depth against zero-shot and free-CoT baselines.We demonstrate that increasing the number of reasoning steps does not consistently improve accuracy across various languages.While highresource and mid-resource languages remain stable, low-resource languages often experience a decline in performance as the number of reasoning steps increases.We attribute this decline to error accumulation and reasoning noise, which are amplified under deeper reasoning in low-resource languages.These findings indicate that CoT is not inherently beneficial, but its effectiveness is significantly influenced by the interaction between reasoning steps and language resource availability.
Authors: Linjian Yang, Xinyan Wang, Kunpeng Liu
Institutions: Clemson University, Portland State University
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Funder: National Science Foundation
Awards: CRII: III: Towards Traceable, Affordable and Explainable Automated Feature Generation for Tabular Data, Collaborative Research: III: Small: Closing Sim-to-Real Gap in Reinforcement Learning via Randomization, Alignment, and Derivation, RII Track-1: AI-enabled Devices for the Advancement of Personalized and Transformative Healthcare in South Carolina
OpenAlex ID: W7166845643