Ziyu Yao's Personal Website
Nguyen Engineering Building, 4415
4400 University Drive
Fairfax, VA, 22030
Contact: ziyuyao at gmu dot edu
Hello! I am an Assistant Professor of the Department of Computer Science at George Mason University, where I co-lead the George Mason NLP group. I am also affiliated with C4I & Cyber Center, Center for Advancing Human-Machine Partnership, and Institute for Digital InnovAtion at GMU. I received my PhD degree from Department of Computer Science and Engineering at the Ohio State University (OSU) in 2021, and have spent time interning at Microsoft Semantic Machines, Carnegie Mellon University, Microsoft Research, Fujitsu Lab of America, and Tsinghua University. My research has been founded by National Science Foundation, Microsoft Accelerate Foundation Models Research Award, Virginia Commonwealth Cyber Initiative, and UMD’s Applied Research Laboratory for Intelligence and Security. I was the Diversity & Inclusion Co-Chair at NAACL 2024 and have organized XLLM-Reason-Plan at COLM 2025, MASC-SLL 2023 (a local NLP event with a 10-year history in the Mid-Atlantic area), SUKI at NAACL 2022, and NLP4Prog at ACL 2021.
With my students, we work on:
- LLMs for Reasoning and Planning: We advance Large Language Models (LLMs) in various tasks demanding reasoning and planning, such as code generation, math reasoning, motion planning, and information extraction.
- Actionable Interpretability of LLMs: We seek to understand the underlying mechanisms of LLMs, particularly applying the interpretation to enhance the model ("actionable interpretability"). Check out our task-centric survey on Mechanistic Interpretability and interpretation of LLMs in code generation and CoT-based arithmetic reasoning.
- Human-AI/LLM Interaction: We study (1) interactive AI/LLM systems, represented by our multi-year effort in "interactive semantic parsing/code generation" -- Preprint'24, ACL'23, ICLR'21, EMNLP'20, EMNLP'19, AAAI'19; (2) problems emerged from real-life deployment of LLMs, such as cost-efficiency of LLM calling and user prompt optimization; and (3) interdisciplinary applications, such as LLM4Edu.
During July 7-11, 2025, we (w/ Dr. Jennifer Suh) organized the first Math EdVenture Summer Camp at GMU, the Fairfax campus, as our commitment to the NSF RITEL project. We hosted a total of 40 middle-school kids across multiple school districts in Northern Virginia. Check out our activities here!
We are organizing a The First Workshop on the Application of LLM Explainability to Reasoning and Planning at COLM 2025! Submit your excellent work to our worshop!
We gave a Tutorial on Mechanistic Interpretability for Language Models at ICML 2025! Our slides are now available.
Excited to release Version 2 of our task-centric survey on Mechanistic Interpretability, in collaboration with Salesforce Research, Purdue University, and George Washington University. Also check out our survey on Sparse Autoencoder (SAE) in collaboration with NJIT and University of Georgia.
**
I plan to recruit 1-3 PhD students for the Fall'26 cycle. In this cycle, the recruit focus is "Actionable Interpretability of LLMs". If you are interested, please email me with your CV, research statement, transcript, future career plan (e.g., academia or industry), representative work, and your reviews on my recent publications, following the instructions on this page!**
**Why George Mason CS?** We rank at No. 35 on CSRanking. We are in one of the BEST locations for studying and living in the US, only 30min driving to Washington DC. I wrote a blog "Why you should apply for Mason CS and work with me" here.
I plan to recruit 1-3 PhD students for the Fall'26 cycle. In this cycle, the recruit focus is "Actionable Interpretability of LLMs". If you are interested, please email me with your CV, research statement, transcript, future career plan (e.g., academia or industry), representative work, and your reviews on my recent publications, following the instructions on this page!** **Why George Mason CS?** We rank at No. 35 on CSRanking. We are in one of the BEST locations for studying and living in the US, only 30min driving to Washington DC. I wrote a blog "Why you should apply for Mason CS and work with me" here.
news
| 08/2025 |  Four papers (2 Main + 2 Findings) were accepted to EMNLP 2025! Thanks for the collaborators! |
|---|---|
| 06/2025 | I gave a talk about “Interpretability Fills the Gap of Data Benchmarks” at the BI4LLMC workshop at FSE! |
| 06/2025 | One paper about VLM reasoning for social robot navigation is accepted to IROS 2025 (w/ Xuesu Xiao and students)! |
| 05/2025 | Two papers accepted to ACL Findings 2025 (Instruction-Tuning for Event Extraction, and Batch Prompting Attack)! Congrats to Murong, Saurabh, and Sweta! |
| 05/2025 | Saurabh, Murong, and Hao will intern at Adobe, Salesforce Research, and Amazon. Mohamed received the CVPR Travel Award. Sai received $1k prize from the CEC Undergrad Research Celebration. Saurabh received the Outstanding GTA Award from GMU CS and was selected to present at the SDM Doctoral Form. Congrats to them!! |
| 04/2025 | Our Tutorial on Mechanistic Interpretability for Language Models is accepted to ICML 2025. Hope to see everyone in Vancouver this summer! |
| 04/2025 | I gave a talk about Language Modeling Reasoning at Georgetown University! Slides |
| 02/2025 | Congrats to Mohamed and collaborators for a paper about Benchmarking VLMs in Path Plan Evaluation being accepted CVPR’25! |
| 01/2025 | Congrats to Murong and collaborators for a paper about “dynamic LLM reasoning” being accepted ICLR’25! |
| 12/2024 | I gave a talk about mechanistic understanding of LLMs in reasoning and code generation at University of Cambridge, Language Technology Lab Seminars! |
| 08/2024 |  Grateful to receive a 3-year grant from National Science Foundation (NSF) on fostering the mathematical modeling competencies of middle-school students by leveraging the power of Generative AI/LLMs (see our prototype)! My first lead-PI NSF project ever;) I’m very lucky to make it with the wonderful Jennifer Suh from GMU Math Edu and Janice Zhang from William & Mary CS! |
| 08/2024 | Congrats to Daking and Saurabh for their two long papers being accepted to ACL 2024 (Main and Findings)! [EDIT] Daking’s paper about mechanistic interpretation of CoT is reported by MIT Technology Review China [English Translate]! |
| 07/2024 |  Check out our latest review on Mechanistic Interpretability for Language Models. We particularly present a taxonomy and a beginner’s roadmap for people interested in this field. |
| 05/2024 | Great pleasure to speak about semantic parsing at the CTO Data Science Speaker Series at Bloomberg NYC! |
| 05/2024 | I was invited to talk about “Interactive Semantic Parsing” at Microsoft Research the PROSE team! |
| 04/2024 | CONGRATS to Murong for being awarded an Outstanding PhD Student by GMU CS, and Hao for receiving the Summer 2024 GRA Fellowship from CAHMP! |
| 04/2024 | Check out our project on “LLM Agents for Mathematics Education” as a teamwork with experts from NLP, HCI, and Mathematics Education! The project was funded by Microsoft Research, Accelerating Foundation Models Research program. |
| 04/2024 | I was invited to talk about “Building Natural Language Interfaces in the Age of LLMs” at Indiana University, Indianapolis! Slides are available here. |
| 03/2024 | I was invited to talk about “Towards Enhancing the Utilization of Large Language Models for Humans” at University of Arizona and Virginia Tech NVC! Slides are available here. |
| 03/2024 | Invited to give a guest lecture about LLMs at Department of Health Administration and Policy and School of Education at GMU! |
| 01/2024 | Our paper on LLM Cascade for Cost Saving in Reasoning (Featured in Hugging Face Daily Papers) is now accepted to ICLR’24! Congrats to Murong and collaborators from Microsoft Research and VT! |
| 12/2023 | Received a grant from Commonwealth Cyber Initiative (CCI) on the topic of LLM supply chain security, with Dr. Xiaokuan Zhang. Thank you, CCI! |
| 11/2023 | I was selected as a Top Reviewer at NeurIPS’23, thanks for the discussions with my students! |
| 10/2023 | Check out our three new preprints on Cost-saving LLMs in reasoning (Featured in Hugging Face Daily Papers), Prompt optimization for zero-shot LLMs, and Benchmarking LLMs in spatial-temporal reasoning! |
| 10/2023 | Congrats to Saurabh and Murong for papers accepted to EMNLP’23 main conference (MailEx, a new conversational event extraction benchmark) and the demo track (Gentopia, our open-source platform streamlining the creation and sharing of augmented LLMs)! Many thanks to our collaborators. See U in Singapore! |
| 09/2023 | Received Azure credits from Microsoft Research, Accelerating Foundation Models Research program for our project on LLM4Edu (in collaboration w/ Dr. Anthony E. Kelly at GMU Educational Psychology)! Thank you, Microsoft! |
| 08/2023 | Received a grant from National Science Foundation (NSF) on the topic of interpretability and explainability of large language models for code! Many thanks to NSF and my collaborators Dr. Kevin Moran (GMU -> UCF) and Dr. Denys Poshyvanyk (W&M)! |
| 08/2023 | I gave in-person talks titled “Building Natural Language Interfaces in the Age of LLMs” at Google and Apple! (Good summer time in the Bay Area  ) |
| 07/2023 |  I was invited to serve as Diversity & Inclusion Chair at NAACL 2024! |
| 06/2023 | Congrats to Hao and Daking for receiving the Graduate Student Travel Fund from GMU! |
| 06/2023 | I was invited to serve as Senior PC member at AAAI’24! |
| 05/2023 | Two papers accepted to ACL’23 (main conference)! Congrats to Hao, Saurabh, and Daking! |
| 04/2023 | George Mason organized MASC-SLL 2023, an annual NLP event in the Mid-Atlantic area! |
| 03/2023 | I was invited to serve as reviewer at NeurIPS’23! |
| 02/2023 | I will be attending AAAI’23 in D.C. and serving as Session Chair! |
| 01/2023 | I was invited to serve as Area Chair at ACL’23 (Question Answering). |
| 12/2022 | Received a grant from Commonwealth Cyber Initiative (CCI) on the topic of algorithm explanation and human trust. Thanks to CCI and my collaborator Dr. Tyler Shaw from GMU Psychology! |
| 11/2022 | Congrats to my student Daking Rai for a student abstract accepted to AAAI! Thanks to my collaborators! |
| 11/2022 | Work about GPT-3 for psychological test item generation got accepted to Journal of Business and Psychology. Thanks to my collaborators! |
| 10/2022 | One paper accepted to EMNLP 2022! |
| 10/2022 | I will be giving a talk in the Department of Statistics at GMU! |
| 09/2022 | I will be giving a (virtual) talk at ServiceNow Research! |
| 08/2022 | I will be attending KDD’22 in person! I will serve as a mentor in the KDD Undergraduate Consortium. |
| 07/2022 | I will be attending NAACL’22 in person! Please join our SUKI workshop in July 14! |
| 07/2022 | I was invited to serve as Senior PC member at AAAI’23. |
| 06/2022 | I was invited to serve as Area Chair at EMNLP’22 (Efficient NLP Track). |
| 04/2022 | Gave a talk in the JHU CLSP seminar! |
| 02/2022 | Consider submitting to the SUKI workshop at NAACL2022! |
| 12/2021 | Our paper “CliniQG4QA” won the Best Paper Award at IEEE BIBM 2021! |
| 11/2021 | Our (w/ Penn State, UW, UCSB, Stanford, UC Berkeley, Google Research) workshop proposal “SUKI: Workshop on Structured and Unstructured Knowledge Integration” was accpted to be co-located with NAACL 2022! Stay tuned! |
| 11/2021 | Invited to serve as ACM SIGAI Newsletter co-editor. |
| 08/2021 | Started my journey at George Mason University! Check out our George Mason NLP group website (co-lead with Prof. Antonios Anastasopoulos)! |
| 05/2021 | I will be interning in Microsoft Semantic Machines in this summer (virtually)! |
| 04/2021 | I am awarded the Graduate Student Research Award by OSU, CSE department! |
| 01/2021 | One paper accepted to ICLR’21! Thanks to my collaborators at CMU! |
| 11/2020 | Super honored to receive the Presidential Fellowship from OSU Graduate School! (“The Presidential Fellowship is the most prestigious award given by the Graduate School. Recipients of this award embody the highest standards of scholarship in the full range of Ohio State’s graduate programs.”) |
| 11/2020 | Our workshop proposal on Natural Language Processing for Programming has been accepted to be co-located with ACL 2021! Congrats to collaborators from Bar-Ilan University, UT Austin, CMU, and OSU! Please stay tuned! |
| 10/2020 | Honored to be selected to the Rising Stars in EECS workshop (hosted by UC Berkeley this year)! |
| 09/2020 | Invited poster at Microsoft Research AI Breakthroughs Workshop (virtually). |
| 08/2020 | Invited talk at VMware, Beijing (virtually). |
| 05/2020 | Excited to start summer internship at CMU Language Technologies Institute with Prof. Graham Neubig! |
| 08/2019 | Our work on a principled Interactive Semantic Parsing framework is accepted to EMNLP! See you in Hong Kong! |
| 07/2019 | Talk at ETH Zurich: “Towards Building Interactive and Collaborative Natural Language Interfaces”. |
| 07/2019 | Attended ACL’19 in Florence. |
| 05/2019 | One paper accepted to ACL’19 (my first ACL paper ever). Congrats to my collaborator Boyuan! |
| 01/2019 | Our work exploring machine collaborations between Code Annotation and Code Retrieval is accepted by WWW’19! |
| 10/2018 | We built an Interactive Semantic Parser: talk to your parser to resolve NL ambiguities (accepted by AAAI’19)! |
| 08/2018 | To know more about StaQC? Check out our work “A Comprehensive Study of StaQC for Deep Code Summarization” (accepted by SIGKDD’18 Deep Learning Day)! |
| 05/2018 | Feeling thrilled to start internship at Microsoft Research @ Redmond this summer!! |
| 04/2018 | Attended WWW 2018 conference @ Lyon, France and present our work “StaQC: A Systematically Mined Question-Code Dataset from Stack Overflow”. Check out the slides and quick StaQC examples! |
| 04/2018 | Attended CRA Grad Cohort Workshop for Women (CRA-W) @ San Francisco. |
| 09/2017 | Gave a talk about “Mining Code Answers to Natural Language Questions” at OSU CSE AI seminar. |
selected publications
2025
- EMNLP’25Feature Extraction and Steering for Enhanced Chain-of-Thought Reasoning in Language ModelsarXiv preprint arXiv:2505.15634 (to appear at EMNLP 2025 Main), 2025
- EMNLP’25All for One: LLMs Solve Mental Math at the Last Token With Information Transferred From Other TokensTo appear at EMNLP 2025 Main, 2025
- EMNLP’25 FindingsA survey on sparse autoencoders: Interpreting the internal mechanisms of large language modelsarXiv preprint arXiv:2503.05613 (to appear at EMNLP 2025 Findings), 2025
- COLM’25WCan LLMs Simulate Personas with Reversed Performance? A Benchmark for Counterfactual Instruction FollowingCOLM Workshop on Social Simulation with LLMs, 2025
- ICLR’25DOTS: Learning to Reason Dynamically in LLMs via Optimal Reasoning Trajectories SearchThe Thirteenth International Conference on Learning Representations, 2025
- AAAI’25WMathVC: An LLM-Simulated Multi-Character Virtual Classroom for Mathematics Education(Invited Presentation at Wolfram Research LLM Agent Colloquium)AAAI AI4Edu Workshop, 2025
2024
- EMNLP’24
- ACL’24An Investigation of Neuron Activation as a Unified Lens to Explain Chain-of-Thought Eliciting Arithmetic Reasoning of LLMs(Covered by MIT Technology Review China [English Translate])ACL, 2024
- ICLR’24Large Language Model Cascades with Mixture of Thoughts Representations for Cost-efficient Reasoning(Featured in Hugging Face Daily Papers)The Twelfth International Conference on Learning Representations (also at ICLR Workshop on Reliable and Responsible Foundation Models), 2024