My name is Zhihan Zhang (张智涵). I am an Applied Scientist at Amazon, where I work on building Rufus, Amazon’s large language model agent tailored for shopping applications. My current work focuses on improving the general instruction-following capabilities of the Rufus model, as part of its post-training efforts.

Prior to joining industry, I earned my Ph.D. in Computer Science from the University of Notre Dame, where I was advised by Dr. Meng Jiang. During my Ph.D., my research centered on training and evaluation methods for instruction-following LLMs. I received my Bachelor’s degree from Peking University, where I worked with Dr. Yunfang Wu and Dr. Xu Sun. I gave a tutorial about instruction-following LLMs at EMNLP 2025.

For my past education and internship experience, please refer to Experience. For the full list of my publications, please refer to Publications or check my Google Scholar page.

News

• Nov 2025: I gave a tutorial about instruction-following LLMs at EMNLP 2025. The tutorial covers training methods, evaluation benchmarks, data collection, and explanability analyses of instruction-following LLMs. Check out our slides and video!

• May 2025: A co-authored paper was accepted by ACL 2025! We proposed a novel framework that leverages implicit user preferences to generate preference tuning data for LLMs.

• May 2025: A co-authored paper was accepted by ACL 2025! We proposed an iterative verify-then-revise framework for LLMs on reasoning tasks.

• Feb 2025: A first-authored paper was accepted by NAACL 2025! We built IHEval, a novel benchmark for assessing LLMs’ capability of following the instruction hierarchy.

• Feb 2025: A co-authored paper was accepted by NAACL 2025! We delivered a new evaluation benchmark for LLM-based recommender systems.

Contact

• Email: zhangzhihan719 [at] gmail.com
• Office: 611 Cowper Street, Palo Alto, CA