Jagle: Building a Large-Scale Japanese Multimodal Post-Training Dataset for Vision-Language Models

• URL: http://arxiv.org/abs/2604.02048v1
• Date: Thu, 02 Apr 2026 13:48:43 GMT
• Title: Jagle: Building a Large-Scale Japanese Multimodal Post-Training Dataset for Vision-Language Models
• Abstract summary: Developing vision-language models (VLMs) that generalize across diverse tasks requires large-scale training datasets with diverse content.<n>In this work, we introduce Jagle, the largest Japanese multimodal post-training dataset to date, comprising approximately 9.2 million instances across diverse tasks.<n>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.
• Score: 27.57801699230882
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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.

Related papers

• Multilingual Training and Evaluation Resources for Vision-Language Models [2.5812054740604453]
  We introduce a new comprehensive suite of resources for Vision Language Models (VLMs) training and evaluation spanning five European languages.<n>We use a regeneration-translation paradigm that produces high-quality cross-lingual resources by combining curated synthetic generation and manual annotation.<n>Experiments show that using multilingual, multimodal examples for training VLMs aids is consistently beneficial on non-English benchmarks.
  arXiv  Detail & Related papers  (2026-04-20T14:42:47Z)
• DEJIMA: A Novel Large-scale Japanese Dataset for Image Captioning and Visual Question Answering [42.08511799479111]
  This work addresses the scarcity of high-quality, large-scale resources for Japanese Vision-and-Language (V&L) modeling.<n>We present a scalable and reproducible pipeline that integrates large-scale web collection with rigorous filtering/deduplication, object-detection-driven evidence extraction, and Large Language Model (LLM)-based refinement.
  arXiv  Detail & Related papers  (2025-11-30T08:09:43Z)
• TowerVision: Understanding and Improving Multilinguality in Vision-Language Models [56.775118098058506]
  TowerVision is a family of open multilingual vision-language models for both image-text and video-text tasks.<n>By incorporating visual and cultural context during fine-tuning, our models surpass existing approaches.<n>To support further research, we publicly release all models, data, and training recipes.
  arXiv  Detail & Related papers  (2025-10-22T17:02:48Z)
• Judging Quality Across Languages: A Multilingual Approach to Pretraining Data Filtering with Language Models [52.22235443948351]
  High-quality multilingual training data is essential for effectively pretraining large language models (LLMs)<n>Here, we introduce JQL, a systematic approach that efficiently curates diverse and high-quality multilingual data at scale.<n>JQL distills LLMs' annotation capabilities into lightweight annotators based on pretrained multilingual embeddings.
  arXiv  Detail & Related papers  (2025-05-28T11:06:54Z)
• Datasets for Multilingual Answer Sentence Selection [59.28492975191415]
  We introduce new high-quality datasets for AS2 in five European languages (French, German, Italian, Portuguese, and Spanish) Results indicate that our datasets are pivotal in producing robust and powerful multilingual AS2 models.
  arXiv  Detail & Related papers  (2024-06-14T16:50:29Z)
• Heron-Bench: A Benchmark for Evaluating Vision Language Models in Japanese [1.2182821723492487]
  We introduce a novel benchmark for evaluating Japanese capabilities of Vision Language Models (VLMs) The Japanese Heron-Bench consists of a variety of imagequestion answer pairs tailored to the Japanese context. We present a baseline Japanese VLM that has been trained with Japanese visual instruction tuning datasets.
  arXiv  Detail & Related papers  (2024-04-11T15:09:22Z)
• RC3: Regularized Contrastive Cross-lingual Cross-modal Pre-training [84.23022072347821]
  We propose a regularized cross-lingual visio-textual contrastive learning objective that constrains the representation proximity of weakly-aligned visio-textual inputs. Experiments on 5 downstream multi-modal tasks across 6 languages demonstrate the effectiveness of our proposed method.
  arXiv  Detail & Related papers  (2023-05-13T14:41:05Z)
• EVJVQA Challenge: Multilingual Visual Question Answering [1.4641199499831683]
  Visual Question Answering (VQA) is a challenging task of natural language processing (NLP) and computer vision (CV) EVJVQA is used as a benchmark dataset for the challenge of multilingual visual question answering at the 9th Workshop on Vietnamese Language and Speech Processing (VLSP 2022) We present details of the organization of the challenge, an overview of the methods employed by shared-task participants, and the results.
  arXiv  Detail & Related papers  (2023-02-23T02:38:39Z)
• Enabling Multimodal Generation on CLIP via Vision-Language Knowledge Distillation [79.72299298976525]
  We propose to augment a vision-language pre-training model with a textual pre-trained language model (PLM) via vision-language knowledge distillation (VLKD) Experiments show that the resulting model has strong zero-shot performance on multimodal generation tasks, such as open-ended visual question answering and image captioning. The original textual language understanding and generation ability of the PLM is maintained after VLKD, which makes our model versatile for both multimodal and unimodal tasks.
  arXiv  Detail & Related papers  (2022-03-12T09:33:37Z)
• UC2: Universal Cross-lingual Cross-modal Vision-and-Language Pre-training [52.852163987208826]
  UC2 is the first machine translation-augmented framework for cross-lingual cross-modal representation learning. We propose two novel pre-training tasks, namely Masked Region-to-Token Modeling (MRTM) and Visual Translation Language Modeling (VTLM) Our proposed framework achieves new state-of-the-art on diverse non-English benchmarks while maintaining comparable performance to monolingual pre-trained models on English tasks.
  arXiv  Detail & Related papers  (2021-04-01T08:30:53Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.