Hitomi Yanaka
Hitomi Yanaka researches language reasoning and explainable AI. Her early papers used natural-deduction proofs to compare sentence meanings and find corresponding phrases. Later shared work tested systematic inference in neural language models, including the SyGNS testbed and Japanese entailment evaluation. Papers from 2024 and 2025 examined language-specific neurons and Transformer behavior during compositional generalization. She directs RIKEN’s Explainable AI Team as of October 2026.
- Now at
- RIKEN and University of Tokyo
- Works
- 109
- Citations
- 288
- h-index
- 8
Publication numbers from OpenAlex, as of 14 October 2026.
Timeline
-
2017–2018
Reasoning with formal proofs
Yanaka coauthored research that used natural-deduction proofs to measure sentence meaning and find corresponding phrases. These papers connected formal reasoning methods with practical language tasks.
-
2020–2022
Testing systematic language understanding
With collaborators, she studied whether neural models learn rules for language inference. The group introduced the SyGNS testbed in 2021 and published a Japanese textual-entailment and similarity evaluation in 2022. The work gave researchers concrete tasks for testing how models generalize.
-
2024–2025
Examining model internals
Yanaka coauthored research on language-specific neurons in multilingual models in 2024 and the internal operations of Transformers in compositional generalization in 2025. These studies connected outward behavior with mechanisms inside the model.
-
As of October 2026
Explainable AI at RIKEN
Yanaka directs RIKEN’s Explainable AI Team. The team studies how AI acquires meaning and performs inference, with the aim of providing explanations that help people make decisions. Its program draws on both humanities and science.
Sources and credits
Photo: www.riken.jp; creator not stated. License: Not stated by source. Source: riken.jp. Cropped to a square and resized. Ask us to remove or replace this photo.