Getting More from Less: Large Language Models are Good Spontaneous Multilingual Learners
This addresses the problem of unbalanced multilingual performance in LLMs for researchers and developers, though it is incremental as it builds on existing translation-based methods.
The study found that instruction-tuning large language models on question translation data, without annotated answers, improves multilingual alignment across a wide range of languages, including unseen ones, demonstrating efficient potential for language and task generalization.
Recently, Large Language Models (LLMs) have shown impressive language capabilities. While most of the existing LLMs have very unbalanced performance across different languages, multilingual alignment based on translation parallel data is an effective method to enhance the LLMs' multilingual capabilities. In this work, we discover and comprehensively investigate the spontaneous multilingual alignment improvement of LLMs. We find that LLMs instruction-tuned on the question translation data (i.e. without annotated answers) are able to encourage the alignment between English and a wide range of languages, even including those unseen during instruction-tuning. Additionally, we utilize different settings and mechanistic interpretability methods to analyze the LLM's performance in the multilingual scenario comprehensively. Our work suggests that LLMs have enormous potential for improving multilingual alignment efficiently with great language and task generalization.