NOW LOADING...

Biography

Ryosuke Takahashi is a first-year Ph.D. student at the Graduate School of Information Sciences, Tohoku University.
He belongs to the Tohoku NLP Group and specializes in natural language processing.

Ryosuke Takahashi
Tohoku NLP Group

Education

Ph.D. of Information Science
Graduate School of Information Sciences, Tohoku University
April 2025  - Present
Master of Information Science
Graduate School of Information Sciences, Tohoku University
April 2023 - March 2025
Bachelor of Engineering
Department of Informatics and Electronics, Faculty of Engineering, Yamagata University
April 2019 - March 2023

Publications

International Conference (Refereed)

Document-level Translation with LLM Reranking: Team-J at WMT 2024 General Translation Task
Keito Kudo*, Hiroyuki Deguchi*, Makoto Morishita*, Ryo Fujii*, Takumi Ito*, Shintaro Ozaki*, Koki Natsumi*, Kai Sato, Kazuki Yano, Ryosuke Takahashi, Subaru Kimura, Tomomasa Hara, Yusuke Sakai, Jun Suzuki (*equal contribution)
In Proceedings of the Ninth Conference on Machine Translation (WMT-2024)
November 2024
Paper
Cite
@inproceedings{kudo-wmt-2024-document,
  
author = {Kudo, Keito and Deguchi, Hiroyuki and Morishita, Makoto and Fujii, Ryo and Ito, Takumi and Ozaki, Shintaro and Natsumi, Koki and Sato, Kai and Yano, Kazuki and Takahashi, Ryosuke and Kimura, Subaru and Hara, Tomomasa and Sakai, Yusuke and Suzuki, Jun},
  
title = {{Document-level Translation with {LLM} Reranking: Team-{J} at {WMT} 2024 General Translation Task}},
  
booktitle = {Proceedings of the Ninth Conference on Machine Translation (WMT)},
  
year = {2024},
  
url = {https://aclanthology.org/2024.wmt-1.14/}
}
The Curse of Popularity: Popular Entities have Catastrophic Side Effects when Deleting Knowledge from Language Models
Ryosuke Takahashi, Go Kamoda, Benjamin Heinzerling, Keisuke Sakaguchi, Kentaro Inui
Non-archival submission for the 17th Annual Conference of the North American Chapter of the Association for Computational Linguistics Student Research Workshop (NAACL SRW 2024)
June 2024
Paper
Cite
@article{takahashi-arxiv-2024-curse,
  
author = {Ryosuke Takahashi and Go Kamoda and Benjamin Heinzerling and Keisuke Sakaguchi and Kentaro Inui},
  
title = {{The Curse of Popularity: Popular Entities have Catastrophic Side Effects when Deleting Knowledge from Language Models}},
  
journal = {arXiv [cs.CL]},
  
year = {2024},
  
url = {https://arxiv.org/abs/2406.06032}
}

Domestic Conference

言語モデルにおける知識の既知性判断の内部表象
佐藤 魁, 高橋 良允, Benjamin Heinzerling, 田中 健史朗, 趙 羽風, 坂井 吉弘, 井之上 直也, 乾 健太郎
第39回人工知能学会全国大会 (JSAI2025)
May 2025
既知性を示す言語表現を伴う知識に関する内部表象の分析
田中 健史朗, 趙 羽風, 坂井 吉弘, 井之上 直也, 佐藤 魁, 高橋 良允, Benjamin Heinzerling, 乾 健太郎
第39回人工知能学会全国大会 (JSAI2025)
May 2025
LMは日本の時系列構造をどうエンコードするか
佐々木睦史, 鴨田豪, 高橋良允, Benjamin Heinzerling, 坂口慶祐
言語処理学会第31回年次大会 (NLP2025)
March 2025
LLM 事前学習の効率化と性質改善: 埋め込み層および出力層のパラメータ固定による再利用
李宰成, 矢野一樹, 高橋良允, 柴田圭悟, 池田航, 鈴木潤
言語処理学会第31回年次大会 (NLP2025)
March 2025
Transformer LLMの内部挙動改善:隠れ状態ベクトルの数値的収束性の向上
柴田圭悟, 高橋良允, 矢野一樹, 李宰成, 池田航, 鈴木潤
言語処理学会第31回年次大会 (NLP2025)
March 2025
Transformer LLM における層単位の FFN 層の重要度検証
池田航, 矢野一樹, 高橋良允, 李宰成, 柴田圭悟, 鈴木潤
言語処理学会第31回年次大会 (NLP2025)
March 2025
LLMにおける内部表現を用いた日本語スタイル制御メカニズムの分析
高橋良允, 矢野一樹, 成瀬健太, 武井美緒, 梶佑輔, 鈴木潤
言語処理学会第31回年次大会 (NLP2025)
March 2025
言語モデルからの知識削除:頻出実体の知識は副作用が破滅的
高橋良允, 鴨田豪 (東北大), Benjamin Heinzerling (理研/東北大), 坂口慶祐 (東北大/理研), 乾健太郎 (MBZUAI/東北大/理研)
言語処理学会第30回年次大会 (NLP 2024)
March 2024
学習済みBERTクラス分類器からのクラス間相関構造の抽出
高橋良允,安田宗樹 (山形大学)
令和4年度情報処理学会東北支部研究会 (IPSJ-TOHOKU)
February 2023

Experience

NLP Department, Mohamed Bin Zayed University of Artificial Intelligence
Visiting Student
November, 2024 - December, 2024
Graduate Program in Data Science (GP-DS),
Tohoku University
Research Assistant
April, 2024 - Present
AI Architecture Section, AI for Business Department,
Rakuten Group, Inc.
Data Scientist (Part-time job)
March, 2024 - Present
AI Agent Group, Data Science & Analytics Department,
Rakuten Group, Inc.
Data Scientist (Intern)
September, 2023 - October, 2023
Biz Freak Inc.
Web Engineer, Data Scientist (Part-time job)
November, 2021 - Present

Awards

日本経済新聞社 CDIO室賞
Paper: LMは日本の時系列構造をどうエンコードするか
NLP2025
(1/765 = 0.1%)
March 2025
若手奨励賞
Paper: 言語モデルからの知識削除:頻出実体の知識は副作用が破滅的
NLP2024
(18/427 = 4.2%)
March 2024
令和4年度 工学部優秀学生賞
Yamagata University
(16/603 = 2.7%)
March 2023
令和4年度 秦逸三記念賞
Yamagata University
(6/603 = 1.0%)
March 2023