Hello, I am Ying Zhang

I am a postdoc in the Natural Language Understanding Team, RIKEN Center for Advanced Intelligence Project (AIP), under the supervision of Prof. Kentaro Inui at Tohoku University.

My current research focuses on the mechanistic interpretability: how language models learn facts during training, store them internally, and use them when answering questions.

ORCID, Semantic Scholar, Google Scholar.

Keywords: Mechanistic Interpretability · Factual Knowledge in Language Models · Fact Recall · Natural Language Processing

My research asks how language models learn facts, where they store them, and how they use them to answer questions. I study these questions by looking inside models and testing how their internal representations change during training and affect their answers..

Our recent work (Special Committee Award, NLP2026) explains how training can enable models to recall learned facts in new queries.

My Ph.D. research focused on improving text generation with re-ranking methods, especially for reducing text degeneration: Addressing Text Degeneration of Discriminative Models with Re-ranking Methods.