WAON：面向对比视觉语言模型文化适配的大规模日文图文数据集

View PDF HTML (experimental)

Abstract:Contrastive vision-language models have achieved remarkable progress through large-scale pretraining. Recent work has shown that removing English-only caption filters and pretraining on global data is effective for improving multicultural performance. We study whether such global pretraining is sufficient for culture-specific understanding, or whether further adaptation with natively sourced data can boost performance beyond what global pretraining alone achieves. To enable this investigation, we present WAON, the largest publicly available native Japanese image-text dataset constructed from native Japanese web content in Common Crawl, containing approximately 155 million examples. We also introduce WAON-Bench, a manually curated Japanese cultural benchmark spanning 374 classes. Through comparative fine-tuning experiments on multiple Japanese image-text datasets, we observe that models fine-tuned on WAON consistently achieve stronger performance on Japanese cultural benchmarks than those fine-tuned on English-to-Japanese translated data. Controlled experiments at matched scale, filtering, and training budget across two model families further indicate that native web origin is the primary driver of this gain. We release our dataset, benchmark, model, and code.

Comments:	Accepted to AACL 2026 (Findings)
Subjects:	Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
Cite as:	arXiv:2510.22276 [cs.CV]
	(or arXiv:2510.22276v4 [cs.CV] for this version)
	https://doi.org/10.48550/arXiv.2510.22276 arXiv-issued DOI via DataCite

Submission history

From: Issa Sugiura [view email]
[v1] Sat, 25 Oct 2025 12:42:42 UTC (2,286 KB)
[v2] Wed, 1 Apr 2026 12:35:54 UTC (8,346 KB)
[v3] Sun, 31 May 2026 10:16:53 UTC (7,562 KB)
[v4] Fri, 2 Oct 2026 02:27:03 UTC (7,750 KB)