Access

TOP    Research    AI for Science Platform Division    AI for Science Foundation Model Research Team

AI for Science Foundation Model Research Team

Japanese

Team Principal  Rio Yokota

  • mail addressrio.yokota [at] riken.jp  (Lab location: Tokyo)
  • Please change [at] to @
2025
Team Principal, AI for Science Foundation Model Research Team, AI for Science Platform Division, RIKEN R-CCS (-present)
2024
PresentProfessor, Institute of Integrated Research, Supercomputing Research Center, Institute of Science Tokyo
2023
Professor, Global Scientific Information and Computing Center, Tokyo Institute of Technology
2015
Associate Professor, Global Scientific Information and Computing Center, Tokyo Institute of Technology
2011
Research Scientist, Applied Mathematics and Computer Science, King Abdullah University of Science and Technology
2010
Post-doctoral Research Associate, Department of Mechanical Engineering, Boston University
2009
Post-doctoral Research Associate, Department of Mathematics, University of Bristol
2009
School of Science for Open and Environmental Systems (PhD), Graduate School of Science and Technology, Keio University

Keyword

  • Ai for Science
  • Foundation Model
  • Reasoning Model
  • Scalable Deep Learning
  • Agentic AI

Research summary

The Advanced General Intelligence for Science Program (TRIP-AGIS) aims to contribute to the dramatic acceleration of the scientific research cycle and the broadening of the scope of exploration within scientific fields, by developing and applying advanced technologies based on artificial superintelligence. The AI for Science Foundation Model Research team aims to develop technology for efficiently training and deploying foundation models with advanced reasoning capabilities, which can assist scientists with tasks such as, proposal of research concepts, planning of experimental programs, development of research code, conducting experiments and simulations, and the analysis of experimental results. To achieve this, we will 1) seek appropriate interfaces and interactions for the foundation models in each scientific field, 2) improve the architecture of the foundation models based on this, 3) develop evaluation benchmarks for scientific domains, and 4) maintain the software infrastructure to enable training and inference to be carried out at maximum speed on next-generation supercomputers.

Representative papers

  • Nakamura, T., Mishra, M., Tedeschi, S., Chai, Y., Stillerman, J. T., Friedrich, F., Yadav, P., Laud, T., Chien, V. M., Zhuo, T. Y. et al.
    Aurora-M: Open Source Continual Pre-training for Multilingual Language and Code
    The 31st International Conference on Computational Linguistics (COLING), Industry Track, Abu Dhabi, UAE, Jan. 2025.
  • Niwa, K., Ishii, H., Sawada, H., Fujino, A., Harada, N., Yokota, R.
    Natural Gradient Primal-Dual Method for Decentralized Learning
    Transactions on Signal and Information Processing over Networks, 10, 417-433, https://doi.org/10.1109/TSIPN.2024.3388948 (2024).
  • Ma, Q, Yokota, R.
    An Inherently Parallel H^2-ULV Factorization for Solving Dense Linear Systems on GPUs
    International Journal of High Performance Computing Applications, 38(4), https://doi.org/10.1177/1094342024124202 (2024).
  • Ootomo, H., Ozaki, K., Yokota, R.
    DGEMM on Integer Matrix Multiplication Unit
    International Journal of High Performance Computing Applications, 38(4), https://doi.org/10.1177/10943420241239588 (2024).
  • Okazaki, N., Hattori, K., Hirai, S., Iida, H., Ohi, M., Fujii, K., Nakamura, T., Loem, M., Yokota, R., Mizuki, S.
    Building a Large Japanese Web Corpus for Large Language Models
    Conference on Language Modeling (COLM), Oct. 2024.
  • Fujii, K., Nakamura, T., Loem, M., Iida, H., Ohi, M., Hattori, K., Hirai, S., Mizuki, S., Yokota, R., Okazaki, N.
    Continual Pre-Training for Cross-Lingual LLM Adaptation: Enhancing Japanese Language Capabilities
    Conference on Language Modeling (COLM), Oct. 2024.
  • Otani, G., Tadokoro, R., Yamada, R., Asano, Y., Laina, I., Rupprecht, C., Inoue, N., Yokota, R., Kataoka, H., Aoki, Y.
    Rethinking Image Super-Resolution from Training Data Perspectives
    European Conference on Computer Vision (ECCV), Sep. 2024.
  • Nakamura, R., Tadokoro, R., Yamada, R., Asano, Y., Laina, I., Rupprecht, C., Inoue, N., Yokota, R., Kataoka, H.
    Scaling Backwards: Minimal Synthetic Pre-training?
    European Conference on Computer Vision (ECCV), Sep. 2024.
  • Yamada, R., Hara, K., Kataoka, H., Makihara, K., Inoue, N., Yokota, R., Satoh, Y.
    Formula-Driven Visual-Geometric Pre-training
    European Conference on Computer Vision (ECCV), Sep. 2024.
  • Shen, Y., Daheim, N., Marconi, G. M., Nickl, P., Cong, B., Clement, B., Yokota, R., Gurevych, I., Cremers, D., Khan, M. E., Möllenhoff, T.
    Variational Learning is Effective for Large Deep Networks
    The 41st International Conference on Machine Learning (ICML), Jul. 2024.

Members

Researcher Search

Related Websites

YOKOTA Laboratory, Supercomputing Research Center, Institute of Integrated Research, Institute of Science TokyoThe webpage will open in a new tab.

Job Openings

Search JobsThe webpage will open in a new tab.