English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Large language models show compelling performance on reasoning tasks but theytend to perform much worse in languages other than English. This isunsurprising given that their training data largely consists of English textand instructions. A typical solution is to translate instruction data into alllanguages of interest, and then train on the resulting multilingual data, whichis called translate-training. This approach not only incurs high cost, but alsoresults in poorly translated data due to the non-standard formatting ofchain-of-thought and mathematical reasoning instructions. In this paper, weexplore the benefits of question alignment, where we train the model totranslate reasoning questions into English by finetuning on X-English questiondata. In this way we perform targetted, in-domain language alignment whichmakes best use of English instruction data to unlock the LLMs' multilingualreasoning abilities. Experimental results on LLaMA2-13B show that questionalignment leads to consistent improvements over the translate-trainingapproach: an average improvement of 11.3\% and 16.1\% accuracy across tenlanguages on the MGSM and MSVAMP maths reasoning benchmarks (The project willbe available at: https://github.com/NJUNLP/QAlign).

Question Translation Training for Better Multilingual Reasoning