sPhinX: Sample Efficient Multilingual Instruction Fine-Tuning Through N-shot Guided Prompting
This addresses the problem of limited multilingual capabilities in LLMs for users needing non-English performance, representing a novel method rather than an incremental improvement.
The paper tackled the performance gap of large language models in non-English languages by introducing sPhinX, a method for constructing multilingual synthetic instruction tuning datasets, and LANGIT, an N-shot guided fine-tuning strategy, which improved Mistral-7B and Phi-3-Small by an average of 39.8% and 11.2% across multilingual benchmarks.
Despite the remarkable success of large language models (LLMs) in English, a significant performance gap remains in non-English languages. To address this, we introduce a novel approach for strategically constructing a multilingual synthetic instruction tuning dataset, sPhinX. Unlike prior methods that directly translate fixed instruction-response pairs, sPhinX enhances diversity by selectively augmenting English instruction-response pairs with multilingual translations. Additionally, we propose LANGIT, a novel N-shot guided fine-tuning strategy, which further enhances model performance by incorporating contextually relevant examples in each training sample. Our ablation study shows that our approach enhances the multilingual capabilities of Mistral-7B and Phi-3-Small improving performance by an average of 39.8% and 11.2%, respectively, across multilingual benchmarks in reasoning, question answering, reading comprehension, and machine translation. Moreover, sPhinX maintains strong performance on English LLM benchmarks while exhibiting minimal to no catastrophic forgetting, even when trained on 51 languages.