This is the code repository of the paper Could Thinking Multilingually Empower LLM Reasoning?.
The code folder contains our code used in the paper, where motivation corresponds to the upper bound analysis part, and methods corresponds to the answer selection strategies part.
The lm-evaluation-harness folder is cloned from the original repo and modified to support our experiments. You can run pip install -e . under the folder to install it.