Ikuya Yamada

Professor, Faculty of Information Science and Technology, Tokyo University of Science
Co-founder & Chief Scientist, Studio Ousia Inc.
Specially Appointed Professor (Visiting), Center for Language AI Research, Tohoku University
Visiting Professor, Center for Artificial Intelligence, Mathematical and Data Science, Nagoya University

I develop trustworthy AI that augments human intelligence, combining fundamental research with real-world applications. My current research explores how LLMs and AI agents can acquire, retrieve, and integrate knowledge from diverse sources to reason, answer questions, and act based on evidence.

I have achieved top results in international AI competitions, including six first-place finishes (details).

ikuya [at] ikuya [dot] net

NLP Laboratory @ TUS 日本語のページ

What's New

July 2026

Co-organized the workshop on Efficient Multimodal Question Answering (EMM-QA) at ICML 2026.

April 2026

Appointed Professor at Tokyo University of Science.

December 2025

Our AI agent won first place in the static evaluation of the Text-to-Text track at the NeurIPS 2025 MMU-RAG Competition.

November 2025

Our paper on KPR, a method for adding knowledge to dense retrievers, was published in Findings of EMNLP 2025.

July 2025

Appointed a Specially Appointed Professor (Visiting) at Tohoku University.

Earlier news

September 2024

Published Introduction to Large Language Models II (in Japanese).

September 2024

Appointed Visiting Professor at Nagoya University.

May 2024

Appointed Senior Action Editor for the ACL Rolling Review.

Research

My research explores how AI acquires and uses knowledge, with a focus on language models, question answering, AI agents, and information extraction. See my Google Scholar profile for a full list of publications.

Language Models

I study how language models incorporate knowledge and use it across languages. We introduced LUKE, a knowledge-enhanced language model; mLUKE, a multilingual model; and LEIA, a method that helps transfer knowledge across languages. The LUKE paper has been cited over 1,000 times, and the model is included in Hugging Face Transformers.

Selected Publications

LUKE: Deep Contextualized Entity Representations with Entity-aware Self-attention (Yamada et al., EMNLP 2020)
mLUKE: The Power of Entity Representations in Multilingual Pretrained Language Models (Ri et al., ACL 2022)
LEIA: Facilitating Cross-lingual Knowledge Transfer in Language Models with Entity-based Data Augmentation (Yamada & Ri, Findings of ACL 2024)

Human–AI quiz match at the NIPS 2017 Human-Computer QA Competition

Question Answering and AI Agents

I study how AI can retrieve relevant information and answer questions based on evidence. Our system won the NIPS 2017 Human-Computer QA Competition and also defeated a team of six U.S. quiz champions. At the NeurIPS 2020 EfficientQA Competition, we placed second in the restricted 6GB track and third in the unrestricted track. At the NeurIPS 2025 MMU-RAG Competition, our AI agent won first place in the static evaluation of the Text-to-Text track. We also introduced KPR, which adds knowledge to dense retrievers without retraining, and BPR, which reduces the memory required for passage retrieval.

Selected Publications

Dynamic Injection of Entity Knowledge into Dense Retrievers (Yamada et al., Findings of EMNLP 2025)
Efficient Passage Retrieval with Hashing for Open-domain Question Answering (Yamada et al., ACL-IJCNLP 2021)
Trick Me If You Can: Human-in-the-Loop Generation of Adversarial Examples for Question Answering (Wallace et al., TACL 2019)

Award ceremony at the WWW 2015 NEEL Challenge

Information Extraction

I study methods for extracting entities and relations from text to build knowledge graphs. We developed Wikipedia2Vec, which represents words and entities in a shared embedding space, as well as entity linking systems that use context to determine which entity a name refers to. We won the WWW 2015 NEEL Challenge, the ACL 2015 W-NUT Shared Task, and ISWC 2020 SemTab. We also placed second in the WSDM Cup 2017 Triple Scoring Task.

Selected Publications

Global Entity Disambiguation with BERT (Yamada et al., NAACL 2022)
Entity Embedding Completion for Wide-Coverage Entity Disambiguation (Oba et al., Findings of EMNLP 2022)
Wikipedia2Vec: An Efficient Toolkit for Learning and Visualizing the Embeddings of Words and Entities from Wikipedia (Yamada et al., EMNLP 2020, System Demonstrations)
Learning Distributed Representations of Texts and Entities from Knowledge Base (Yamada et al., TACL 2017)
Joint Learning of the Embedding of Words and Entities for Named Entity Disambiguation (Yamada et al., CoNLL 2016)

Books

I co-authored and supervised books in Japanese that bridge theory and practice in natural language processing.

Introduction to Large Language Models II (in Japanese)
Ikuya Yamada (Author & Supervisor), Masatoshi Suzuki, Sosuke Nishikawa, Kazuki Fujii, Kosuke Yamada, Ryokan Ri
Gijutsu-Hyoron Co., Ltd., 2024
Introduction to Large Language Models (in Japanese)
Ikuya Yamada (Author & Supervisor), Masatoshi Suzuki, Kosuke Yamada, Ryokan Ri
Gijutsu-Hyoron Co., Ltd., 2023
Natural Language Processing with Deep Learning (in Japanese)
Ikuya Yamada, Tomohide Shibata, Hiroyuki Shindo, Ryuji Tamaki
Kyoritsu Shuppan Co., Ltd., 2023
Natural Language Processing (in Japanese)
Jun Suzuki, Seiji Tsuchiya, Kazuki Motohashi, Kanji Takahashi, Akihiro Tamura, Ikuya Yamada, Kenji Araki, Shinsuke Mori, Shinichi Watanabe, Shin Hara, Tomoya Mizumoto, Takeshi Shimizu
Johokiko Co., Ltd., 2020
Introduction to Machine Learning Programming in Python (in Japanese)
Tatsuro Shimada, Naoto Koshimizu, Atsushi Hayakawa, Ikuya Yamada
Gijutsu-Hyoron Co., Ltd., 2019

Awards

1st Place: NeurIPS 2025 MMU-RAG Competition (Text-to-Text track, static evaluation)
2nd Place: NeurIPS 2020 Efficient Open-Domain Question Answering Competition (restricted 6GB track)
3rd Place: NeurIPS 2020 Efficient Open-Domain Question Answering Competition (unrestricted track)
1st Place: ISWC 2020 Semantic Web Challenge on Tabular Data to Knowledge Graph Mining
1st Place: NIPS 2017 Human-Computer Question Answering Competition
2nd Place: WSDM Cup 2017 Triple Scoring Task
1st Place: Shared Task in NAACL 2016 Workshop on Human-Computer QA
1st Place: Shared Task #1 in ACL 2015 Workshop on Noisy User-generated Text (W-NUT 2015)
1st Place: NEEL Challenge in WWW 2015 Workshop on Making Sense of Microposts

Career

I founded Newrong Inc. upon entering university and, following its acquisition by Fractalist Inc., went on to co-found Studio Ousia Inc. I am now a Professor at Tokyo University of Science and Chief Scientist at Studio Ousia, with visiting appointments at Tohoku University and Nagoya University.

2000: Graduated from Keio Shonan Fujisawa Senior High School
2000: Appointed CEO of Newrong Inc.
2003: Appointed Director at Fractalist Inc. (following the acquisition of Newrong Inc.)
2005: Graduated from the Faculty of Environment and Information Studies, Keio University
2007: Appointed Representative Director and Chief Scientist at Studio Ousia Inc.
2011: Completed the master’s program at the Graduate School of Media and Governance, Keio University
2016: Earned a Ph.D. from the Graduate School of Media and Governance, Keio University
2018: Appointed Visiting Researcher at the RIKEN Center for Advanced Intelligence Project
2023: Appointed Chief Scientist at Studio Ousia Inc. (present)
2024: Appointed Visiting Professor at the Center for Artificial Intelligence, Mathematical and Data Science, Nagoya University (present)
2025: Appointed a Specially Appointed Professor (Visiting) at the Center for Language AI Research, Tohoku University (present)
2026: Appointed Professor in the Faculty of Information Science and Technology, Tokyo University of Science (present)