Ziyu Yao
Scholar

Ziyu Yao

Google Scholar ID: 4lYrMNUAAAAJ
Assistant Professor, George Mason University
LLM Reasoning/PlanningMechanistic InterpretabilityHuman-LLM Interaction
Citations & Impact
All-time
Citations
1,496
 
H-index
19
 
i10-index
24
 
Publications
20
 
Co-authors
22
list available
Resume (English only)
Academic Achievements
  • Research funded by the National Science Foundation (NSF), Microsoft Accelerate Foundation Models Research Award, and Virginia Commonwealth Cyber Initiative
  • Served as Diversity & Inclusion Co-Chair at NAACL 2024
  • Delivered a tutorial on 'Mechanistic Interpretability for Language Models' at ICML 2025
  • Released Version 2 of a task-centric survey on Mechanistic Interpretability in collaboration with Salesforce Research, Purdue University, and George Washington University
  • Published a survey on Sparse Autoencoders (SAE) with NJIT and the University of Georgia
  • Paper 'Failure by Interference' received two best paper nominations and a Best Paper Honorable Mention at the First Workshop on the Application of LLM Explainability to Reasoning and Planning
Research Experience
  • Assistant Professor at George Mason University, leading the NLP research group
  • Interned at Microsoft Semantic Machines, Carnegie Mellon University, Microsoft Research, Fujitsu Lab of America, and Tsinghua University
  • Co-organized the Math EdVenture Summer Camp at GMU in July 2025 with Dr. Jennifer Suh as part of the NSF RITEL project
  • Organized multiple workshops including XLLM-Reason-Plan at COLM 2025, MASC-SLL 2023, SUKI at NAACL 2022, and NLP4Prog at ACL 2021
  • Plans to recruit 2–3 PhD students for Fall 2026, with a focus on 'Actionable Interpretability of LLMs'
Background
  • Assistant Professor in the Department of Computer Science at George Mason University
  • Co-leads the George Mason NLP group
  • Research focuses on: LLMs for reasoning and planning, actionable interpretability of LLMs, and human-AI/LLM interaction
  • Specific tasks include code generation, math reasoning, motion planning, and information extraction
  • Aims to apply LLM interpretability to real-world systems and explore interdisciplinary applications such as LLM4Edu