X-LLaVA: Optimizing Bilingual Large Vision-Language Alignment
This work addresses the problem of expensive and challenging multilingual data construction for multimodal models, offering incremental improvements for researchers and developers in multilingual AI.
The paper tackles the high cost and complexity of creating multilingual multimodal training data by proposing two cost-effective methods, constructing a 91K English-Korean-Chinese dataset, and developing a bilingual model that outperforms existing approaches in Korean and English.
The impressive development of large language models (LLMs) is expanding into the realm of large multimodal models (LMMs), which incorporate multiple types of data beyond text. However, the nature of multimodal models leads to significant expenses in the creation of training data. Furthermore, constructing multilingual data for LMMs presents its own set of challenges due to language diversity and complexity. Therefore, in this study, we propose two cost-effective methods to solve this problem: (1) vocabulary expansion and pretraining of multilingual LLM for specific languages, and (2) automatic and elaborate construction of multimodal datasets using GPT4-V. Based on015 these methods, we constructed a 91K English-Korean-Chinese multilingual, multimodal training dataset. Additionally, we developed a bilingual multimodal model that exhibits excellent performance in both Korean and English, surpassing existing approaches.