Is continuous CoT better suited for multi-lingual reasoning?
This work addresses the problem of improving cross-lingual reasoning for multilingual NLP models, particularly for low-resource languages, by leveraging continuous latent representations.
This paper investigates whether continuous latent space reasoning improves multilingual capabilities, comparing Continuous Chain-of-Thought (CODI) against standard fine-tuning across five languages on GSM8k and CommonsenseQA. The study found that continuous reasoning significantly outperforms explicit reasoning in low-resource languages, especially in zero-shot settings, while also achieving high efficiency by compressing reasoning traces by 29x to 50x.
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.