From February 18 – March 19, 2026, Prof. Michael Färber spent one month in Tokyo, Japan for a research stay at the National Institute of Informatics (NII). The stay was hosted by Prof. Akiko Aizawa (NII). It centered on current research at the intersection of trustworthy AI, language models for science, retrieval, and structured knowledge. The topical focus was on:

• evidence-grounded generation with citations,
• retrieval-augmented generation,
• the integration of structured knowledge such as knowledge graphs into language-model pipelines, and
• broader questions of evaluation, transparency, and reliability of large language models.

Program

During his time in Tokyo, Prof. Färber joined a variety of meetings and regular exchanges with Prof. Aizawa’s group and beyond. The research stay had a strong working character, combining invited talks, lab visits, and research discussions across several leading AI research environments in Japan.

To kick off his research stay, Prof. Färber gave a talk to Prof. Aizawa’s group at NII on February 20, 2026. This provided an opportunity for in-depth exchange with researchers from the group. On the same day, he met with Assistant Professor Chifumi Nishioka.

Next, Prof. Färber gave a guest talk at the Institute of Science Tokyo. Under the title “From papers to insights with LLMs & Knowledge Graphs”, he shared insights from his recent work in this direction. The visit included a presentation as well as an extended research-focused conversation with the group of Prof. Naoaki Okazaki.

At the University of Tokyo, Prof. Färber visited the Matsuo-Iwasawa Laboratory. There, he discussed his current work on large-scale AI systems and language models with the research group.

He also exchanged with researchers at the NII LLM Center. Those meetings focused in particular on LLM safety, evaluation, and reliability. The discussions connected closely to ongoing work at ScaDS.AI Dresden/Leipzig on trustworthy AI for science.

Additional meetings were held with Prof. Makoto P. Kato (NII/University of Tsukuba) and with Prof. Saku Sugawara and his group. These meetings provided further perspectives in the direction of information retrieval and its link to evidence-grounded generation, evaluation, and language understanding.

Outlook

During his research stay, Prof. Färber strengthened existing collaborations, created new contacts, and opened concrete directions for future joint work between ScaDS.AI Dresden/Leipzig and the partners in Japan. At the same time, the stay increased the international visibility of TU Dresden and ScaDS.AI Dresden/Leipzig in the field of trustworthy AI, while highlighting the value of long-term research cooperations between Germany and Japan. More information on Prof. Färber’s work and chair is available here.

JSPS BRIDGE Fellowship

Prof. Färber’s research stay was supported by a JSPS BRIDGE Fellowship. The program enables former JSPS fellows to return to Japan to deepen existing collaborations and build new long-term international academic connections. Learn more about the JSPS BRIDGE Fellowship program here.