Koki Maeda

JaWildText: A Benchmark for Vision-Language Models on Japanese Scene Text Understanding

arXiv preprint · 2026

A diagnostic benchmark for evaluating Japanese scene text understanding in vision-language models across STVQA, receipt KIE, and handwriting OCR.

Koki Maeda, Naoaki Okazaki

JaWildText: 日本語文字認識性能評価のための実世界画像データセット

言語処理学会第32回年次大会 (NLP2026) · 2026

日本語の実世界画像における文字認識と下流タスクを一体的に評価するためのデータセット JaWildText を提案する。高密度 STVQA、レシート KIE、手書き文字 OCR の 3 つの課題を収録し、公開 VLM および OCR 特化モデルの評価を通じて、日本語読字性能になお改善余地があることを示した。NLP2026 委員特別賞受賞論文。

前田 航希, 岡崎 直観

From Correspondence to Actions: Human-Like Multi-Image Spatial Reasoning in Multi-modal Large Language Models

arXiv preprint · 2026

Proposes a framework for evaluating human-like multi-image spatial reasoning in multi-modal large language models, analyzing how models establish correspondence across images and translate spatial understanding into actions.

Masanari Ohi, Koki Maeda, Ryuto Koike, Daisuke Oba, Nakamasa Inoue, Naoaki Okazaki

Why We Build Local Large Language Models: An Observational Analysis from 35 Japanese and Multilingual LLMs

The 1st Workshop on Multilingual and Equitable Language Technologies (MELT) · 2025

Koshiro Saito, Sakae Mizuki, Masanari Ohi, Taishi Nakamura, Taihei Shiotani, Koki Maeda, Youmi Ma, Kakeru Hattori, Kazuki Fujii, Takumi Okamoto, Shigeki Ishida, Hiroya Takamura, Rio Yokota, Naoaki Okazaki

Building instruction-tuning datasets from human-written instructions with open-weight large language models

The 2nd Conference on Language Modeling (COLM) · 2025

Youmi Ma, Sakae Mizuki, Kazuki Fujii, Taishi Nakamura, Masanari Ohi, Hinari Shimada, Taihei Shiotani, Koshiro Saito, Koki Maeda, Kakeru Hattori, Takumi Okamoto, Shigeki Ishida, Rio Yokota, Hiroya Takamura, Naoaki Okazaki