Is continuous CoT better suited for multi-lingual reasoning?
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914379853725696 |
|---|---|
| author | Bashir, Ali Hamza Shomali, Behzad Frey, Markus Ali, Mehdi Sifa, Rafet Berghaus, David |
| author_facet | Bashir, Ali Hamza Shomali, Behzad Frey, Markus Ali, Mehdi Sifa, Rafet Berghaus, David |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_08177 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Is continuous CoT better suited for multi-lingual reasoning? Bashir, Ali Hamza Shomali, Behzad Frey, Markus Ali, Mehdi Sifa, Rafet Berghaus, David Computation and Language Artificial Intelligence Machine Learning 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. |
| title | Is continuous CoT better suited for multi-lingual reasoning? |
| topic | Computation and Language Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2603.08177 |