Is continuous CoT better suited for multi-lingual reasoning?

TL;DR

This paper shows that continuous latent space reasoning improves multilingual reasoning robustness and efficiency, especially for low-resource languages, by leveraging language-invariant representations in a scalable manner.

Contribution

It introduces the use of continuous Chain-of-Thought reasoning with the CODI framework for multilingual tasks, demonstrating superior performance and efficiency over explicit reasoning methods.

Findings

Continuous reasoning outperforms explicit reasoning in low-resource languages.

The approach achieves 29-50 times compression of reasoning traces.

Continuous latent representations exhibit greater language invariance.

Abstract

We investigate whether performing reasoning in a continuous latent space leads to more robust multilingual capabilities. We compare Continuous Chain-of-Thought (using the CODI framework) against standard supervised fine-tuning across five typologically diverse languages: English, Chinese, German, French, and Urdu. Our experiments on GSM8k and CommonsenseQA demonstrate that continuous reasoning significantly outperforms explicit reasoning on low-resource languages, particularly in zero-shot settings where the target language was not seen during training. Additionally, this approach achieves extreme efficiency, compressing reasoning traces by approximately $29\times$ to $50\times$. These findings indicate that continuous latent representations naturally exhibit greater language invariance, offering a scalable solution for cross-lingual reasoning.