New Japanese image-text dataset boosts AI cultural understanding

Researchers have introduced WAON, a large-scale Japanese image-text dataset comprising approximately 155 million examples sourced from native Japanese web content. This dataset aims to improve the cultural understanding of contrastive vision-language models. Alongside WAON, they developed WAON-Bench, a curated benchmark for Japanese cultural understanding with 374 classes. Experiments show that models fine-tuned on WAON outperform those trained on translated English data for Japanese cultural tasks. AI

IMPACT Enables development of AI models with improved understanding of Japanese culture and language nuances.

RANK_REASON The cluster describes a new academic paper introducing a dataset and benchmark for AI research. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• Common Crawl
• Contrastive vision-language models
• Issa Sugiura
• WAON-Bench

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →