Language Models are Multilingual Chain-of-Thought Reasoners
This work addresses the need for multilingual reasoning benchmarks for AI researchers, though it is incremental as it builds on existing datasets and methods.
The paper tackles the problem of evaluating multilingual reasoning in large language models by introducing the MGSM benchmark, translating 250 math problems into ten languages, and finds that chain-of-thought prompting improves with scale, showing strong abilities even in underrepresented languages like Bengali and Swahili.
We evaluate the reasoning abilities of large language models in multilingual settings. We introduce the Multilingual Grade School Math (MGSM) benchmark, by manually translating 250 grade-school math problems from the GSM8K dataset (Cobbe et al., 2021) into ten typologically diverse languages. We find that the ability to solve MGSM problems via chain-of-thought prompting emerges with increasing model scale, and that models have strikingly strong multilingual reasoning abilities, even in underrepresented languages such as Bengali and Swahili. Finally, we show that the multilingual reasoning abilities of language models extend to other tasks such as commonsense reasoning and word-in-context semantic judgment. The MGSM benchmark is publicly available at https://github.com/google-research/url-nlp.