Hi! I'm Issa Sugiura

Research Interests

My research focuses on building vision-language models (VLMs) with strong Japanese language and cultural understanding. To get there, I work across the full development pipeline — building training datasets and benchmarks, and developing the Japanese models themselves:

• constructing large-scale training datasets (e.g., WAON, Jagle)
• developing reliable evaluation benchmarks (e.g., JAMMEval, HakushoBench)
• building Japanese-centric VLMs (e.g., LLM-jp-4-VL 9B beta)

I am also interested in benchmarking LLM agents — particularly on economically valuable tasks that go beyond coding and math, and on long-horizon tasks (e.g., CoffeeBench, EDINET-Bench).

Education

• Apr 2026 – present

  Doctoral Program (Expected)

  Institute of Science Tokyo, Japan

  Course of Artificial Intelligence, Department of Computer Science, School of Computing

• Apr 2024 – Mar 2026

  Master's Degree

  Kyoto University, Japan

  Course of Communications and Computer Engineering, Graduate School of Informatics

• Apr 2020 – Mar 2024

  Bachelor's Degree

  Osaka University, Japan · GPA 3.57/4.00

  Software Science Course, Department of Information and Computer Sciences, School of Engineering Science

Experience

• Apr 2026 – present

  Research Assistant @ Institute of Science Tokyo

  Research on VLMs.

• Dec 2024 – present

  Student Intern @ Sakana AI

  Building benchmarks to evaluate LLM agents on economically valuable, long-horizon tasks, including financial analysis (EDINET-Bench, ICLR 2026) and multi-agent economies (CoffeeBench).

• Feb 2024 – present

  Research Assistant @ National Institute of Informatics

  Developing Japanese vision-language models across the full pipeline — large-scale dataset construction (WAON, Jagle), evaluation benchmark design (JAMMEval, HakushoBench), and model training (LLM-jp-4-VL).

Publications

Peer-Reviewed Publications

• 1.

  CoffeeBench: Benchmarking Long-Horizon LLM Agents in Heterogeneous Multi-Agent Economies

  Issa Sugiura, Daichi Hattori, Kazuo Araragi, Keita Ogawa, Shota Onose, Taro Makino, Teppei Usuki, Takashi Ishida

  ICML 2026 Workshop on Failure Modes in Agentic AI, July 2026

  Paper Code Blog
• 2.

  Cross-Task Evaluation and Empirical Analysis of Japanese Visual Language Models

  Koki Maeda, Issa Sugiura, Yusuke Oda, Shuhei Kurita, Naoaki Okazaki

  Journal of Natural Language Processing, June 2026

  Paper Code
• 3.

  EDINET-Bench: Evaluating LLMs on Complex Financial Tasks using Japanese Financial Statements

  Issa Sugiura, Takashi Ishida, Taro Makino, Chieko Tazuke, Takanori Nakagawa, Kosuke Nakago, David Ha

  ICLR 2026, Apr 2026

  Paper Code Dataset
• 4.

  Removing Mislabeled Data from Trained Models via Machine Unlearning

  Issa Sugiura, Shingo Okamura, Naoto Yanai

  IEICE Transactions on Information and Systems, Aug 2025

  Paper Code
• 5.

  Developing Japanese CLIP Models Leveraging an Open-weight LLM for Large-scale Dataset Translation

  Issa Sugiura, Shuhei Kurita, Yusuke Oda, Daisuke Kawahara, Naoaki Okazaki

  NAACL 2025 Student Research Workshop, Apr 2025

  Paper Code Dataset Model
• 6.

  Constructing Multimodal Datasets from Scratch for Rapid Development of a Japanese Visual Language Model

  Keito Sasagawa, Koki Maeda, Issa Sugiura, Shuhei Kurita, Naoaki Okazaki, Daisuke Kawahara

  NAACL 2025 Demo Track, Apr 2025

  Paper Model
• 7.

  A Comprehensive Analysis of Memorization in Large Language Models

  Hirokazu Kiyomaru*, Issa Sugiura*, Daisuke Kawahara, Sadao Kurohashi (*equal contribution)

  INLG 2024, Sep 2024

  Paper Code

Preprints

• 1.

  HakushoBench: A Japanese Chart and Table VQA Benchmark from Governmental White Papers

  Issa Sugiura, Shuhei Kurita, Yusuke Oda, Naoaki Okazaki

  arXiv, Jun 2026

  Paper Code Dataset
• 2.

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

  Issa Sugiura, Keito Sasagawa, Keisuke Nakao, Koki Maeda, Ziqi Yin, Zhishen Yang, Shuhei Kurita, Yusuke Oda, Ryoko Tokuhisa, Daisuke Kawahara, Naoaki Okazaki

  arXiv, Apr 2026

  Paper Code Dataset Model
• 3.

  JAMMEval: A Refined Collection of Japanese Benchmarks for Reliable VLM Evaluation

  Issa Sugiura, Koki Maeda, Shuhei Kurita, Yusuke Oda, Daisuke Kawahara, Naoaki Okazaki

  arXiv, Apr 2026

  Paper Code Dataset
• 4.

  WAON: Large-Scale and High-Quality Japanese Image-Text Pair Dataset for Vision-Language Models

  Issa Sugiura, Shuhei Kurita, Yusuke Oda, Daisuke Kawahara, Yasuo Okabe, Naoaki Okazaki

  arXiv, Oct 2025

  Paper Code WAON WAON-Bench
• 5.

  Llama-Mimi: Speech Language Models with Interleaved Semantic and Acoustic Tokens

  Issa Sugiura, Shuhei Kurita, Yusuke Oda, Ryuichiro Higashinaka

  arXiv, Sep 2025

  Paper Code Model Demo
• 6.

  llm-jp-modernbert: A ModernBERT Model Trained on a Large-Scale Japanese Corpus with Long Context Length

  Issa Sugiura, Kouta Nakayama, Yusuke Oda

  arXiv, Apr 2025

  Paper Code Model
• 7.

  LLM-jp: A Cross-organizational Project for the Research and Development of Fully Open Japanese LLMs

  LLM-jp (incl. Issa Sugiura; 80 authors, listed in alphabetical order)

  arXiv, Jul 2024

  Paper

Domestic Conference (Non-Refereed)

• 1.

  Jagle: 視覚言語モデルのための大規模日本語マルチモーダル事後学習データセットの構築

  杉浦一瑳, 笹川慶人, 中尾圭佑, 前田航希, Yin Ziqi, Yang Zhishen, 栗田修平, 小田悠介, 徳久良子, 河原大輔, 岡崎直観

  JSAI2026, 2026年6月

  Paper Project Page Poster
• 2.

  Llama-Mimi: 意味・音響トークンを交互配置した音声言語モデル

  最優秀賞

  杉浦一瑳, 栗田修平, 小田悠介, 東中竜一郎

  NLP2026, 2026年3月

  Paper Code Model Demo
• 3.

  JAMMEval: 再アノテーションによる日本語VQA評価データセットの信頼性向上

  杉浦一瑳, 前田航希, 栗田修平, 小田悠介, 河原大輔, 岡崎直観

  NLP2026, 2026年3月

  Paper
• 4.

  WAON: 視覚言語モデルのための大規模かつ高品質な日本語画像・テキスト対データセット

  委員特別賞

  杉浦一瑳, 栗田修平, 小田悠介, 河原大輔, 岡部寿男, 岡崎直観

  NLP2026, 2026年3月

  Paper Project Page
• 5.

  Common Crawlを用いた大規模音声音響データセットの構築

  浅井航平, 杉浦一瑳, 中田亘, 栗田修平, 高道慎之介, 小川哲司, 東中竜一郎

  日本音響学会 2025年秋季研究発表会, 2025年9月

  Dataset
• 6.

  オープンLLMによる翻訳を活用した日本語CLIPの開発

  杉浦一瑳, 栗田修平, 小田悠介, 河原大輔, 岡崎直観

  NLP2025, 2025年3月

  Paper Code Dataset Model
• 7.

  ロススパイクの影響分析

  杉浦一瑳, 栗田修平, 小田悠介

  NLP2025, 2025年3月

  Paper
• 8.

  llm-jp-eval-mm: 日本語視覚言語モデルの自動評価基盤

  若手奨励賞

  前田航希*, 杉浦一瑳*, 栗田修平, 小田悠介, 岡崎直観

  NLP2025, 2025年3月

  Paper Code
• 9.

  LLM-jp-3 VILA: 日本語マルチモーダルデータセット及び強力な日本語マルチモーダルモデルの構築

  委員特別賞

  笹川慶人, 前田航希, 杉浦一瑳, 栗田修平, 岡崎直観, 河原大輔

  NLP2025, 2025年3月

  Paper
• 10.

  ミスラベルデータの忘却による学習済みモデルの汎化性能の向上手法の提案

  LINEヤフースポンサー賞

  杉浦一瑳, 岡村真吾, 山下恭佑, 矢内直人

  DEIM2024, 2024年2月

  Paper Code Slide

Models / Systems

• 1.

  LLM-jp-4-VL 9B beta (2026)

  Developed a 9B Japanese vision-language model achieving competitive performance on Japanese benchmarks.

  Model Blog

Certifications / Qualifications

• Oct 19, 2025 TOEIC L&R: L 420 + R 425 = 845
• Aug 2022 Security Camp organized by IPA (Information-technology Promotion Agency), Web Security Course