Michihiro Yasunaga

Michihiro Yasunaga is an Assistant Professor in the Department of Computer Science at Stanford University's School of Engineering. He received his PhD in Computer Science from Stanford, advised by Percy Liang, Jure Leskovec, and Chris Manning. Prior to his faculty position, he worked as a researcher at Google DeepMind and Meta.

His research focuses on building LLMs and agents that assist humans in diverse tasks, with particular expertise in post-training techniques (RL, reward models, and evaluation), reasoning systems (AnalogicalReasoner), retrieval and tool use for LLMs (LinkBERT, QAGNN, DRAGON, REPLUG, HippoRAG), and multimodality (RA-CM3, Med-Flamingo, Transfusion). His work spans both theoretical foundations and practical applications of large language models.

Yasunaga's publication record demonstrates significant contributions to the field of AI, with 15 recent articles (2023-2025) covering diverse aspects of language model development, evaluation, and application. His research shows a clear trajectory toward building more capable, efficient, and reliable multimodal AI systems, with particular emphasis on knowledge integration and robust evaluation frameworks.

Among his notable achievements is the Best Paper Award at AAAI 2023 Deep Learning on Graphs Workshop for the DRAGON paper. He has also been deeply involved in major benchmarking efforts including HELM and HEIM, which provide comprehensive evaluation frameworks for language and vision-language models.

Yasunaga actively contributes to the research community through service roles including Organizing Committee for the Workshop on Knowledge-Augmented Methods for NLP (ACL 2024), Workshop on Structured and Unstructured Knowledge Integration (NAACL 2022), and the Workshop on Scientific Document Summarization (SIGIR 2017-2020). He has also served on program committees for top conferences including NeurIPS, ICML, ICLR, ACL, EMNLP, CVPR, and ICCV from 2020-2025.