Jagle: Building a Large-Scale Japanese Multimodal Post-Training Dataset for Vision-Language Models
This addresses the problem of limited training data for building high-quality multilingual vision-language models, particularly for Japanese, and is incremental as it extends existing dataset construction methods to a new language.
The authors tackled the lack of large-scale multimodal datasets for non-English languages by creating Jagle, a Japanese dataset with 9.2 million instances, which improved a 2.2B model's performance on Japanese tasks, surpassing InternVL3.5-2B and nearing Qwen3-VL-2B-Instruct, while also enhancing English performance when combined with FineVision.
Developing vision-language models (VLMs) that generalize across diverse tasks requires large-scale training datasets with diverse content. In English, such datasets are typically constructed by aggregating and curating numerous existing visual question answering (VQA) resources. However, this strategy does not readily extend to other languages, where VQA datasets remain limited in both scale and domain coverage, posing a major obstacle to building high-quality multilingual and non-English VLMs. In this work, we introduce Jagle, the largest Japanese multimodal post-training dataset to date, comprising approximately 9.2 million instances across diverse tasks. Rather than relying on existing VQA datasets, we collect heterogeneous source data, including images, image-text pairs, and PDF documents, and generate VQA pairs through multiple strategies such as VLM-based QA generation, translation, and text rendering. Experiments demonstrate that a 2.2B model trained with Jagle achieves strong performance on Japanese tasks, surpassing InternVL3.5-2B in average score across ten Japanese evaluation tasks and approaching within five points of Qwen3-VL-2B-Instruct. Furthermore, combining Jagle with FineVision does not degrade English performance; instead, it improves English performance compared to training with FineVision alone. To facilitate reproducibility and future research, we release the dataset, trained models, and code.