|
Zhuofan Jia Pronounced as JWOH-fahn JYAH Hi, I am a third-year PhD student in Computer Science at Duke University, advised by Prof. Jian Pei. I received my bachelor's degrees in Computer Science and Mathematics from University of Illinois Urbana-Champaign, where I worked with Prof. Gang Wang. My research centers on developing a principled understanding of large language models and data-driven machine learning, with the goal of making these models more reliable, trustworthy, and useful in real-world settings. Email / CV / Google Scholar / LinkedIn |
|
ResearchI am particularly interested in how data shapes model behavior during training and inference, including evaluation and benchmarking for complex conversational settings, privacy and unlearning in language models, and uncertainty-aware methods for data valuation and model assessment. |
|
MPCEval: A Benchmark for Multi-Party Conversation Generation
Minxing Zhang, Yi Yang, Zhuofan Jia, Xuan Yang, Jian Pei, Yuchen Zang, Xingwang Deng, Xianglong Chen KDD, 2026 arXiv A task-aware benchmark for multi-party conversation generation that evaluates both next-turn prediction and full-conversation quality with reproducible, reference-free metrics for speaker behavior, content quality, and their consistency. |
|
Shapley Value on Uncertain Data
Zhuofan Jia, Jian Pei Under Review (TKDE), 2026 arXiv A probabilistic data-Shapley framework that models each participant's value as a random variable induced by sampling, and estimates both its expectation and variance with unbiased and variance-efficient Monte Carlo methods. |
|
VeriSMS: A Message Verification System for Inclusive Patient Outreach against Phishing Attacks
Chenkai Wang, Zhuofan Jia, Hadjer Benkraouda, Cody Zevnik, Nicholas Heuermann, Roopa Foulger, Jonathan A. Handler, Gang Wang Proceedings of ACM CHI Conference on Human Factors in Computing Systems (CHI), 2024 paper An inclusive patient-outreach verification system that lets users confirm healthcare messages through SMS and phone calls, improving phishing detection without requiring smartphones or extra setup. |
Selected Projects |
|
Fine-Tuning and Evaluation of Small LLMs
Large Language Models A fine-tuned modular Gemma-3-270M system that combines task routing, supervised fine-tuning, knowledge distillation, and retrieval augmentation to improve factual QA, reasoning, and instruction following. |
|
Beyond Binary Truth: Benchmarking the Rigor of LLM Agents in Scientific Hypothesis Verification
Large Language Model Agents A benchmark for false-hypothesis rejection that uses controlled claim perturbations to evaluate whether tool-using scientific agents can rigorously reject invalid hypotheses through executable, data-grounded analysis. |
Professional Experience |
|
Applied Scientist Intern
Amazon 路 AME Outbound Team Seattle, WA 路 May 2026 - Present |
|
Software Engineer Intern
HERE Technologies 路 Dynamic Content Team Chicago, IL 路 Jan 2023 - May 2023 Built machine learning solutions for location intelligence, including store-status prediction models with F1 above 95% and privacy-preserving geospatial data processing for customer trip data. |
|
Machine Learning Engineer Intern
Discovery Partners Institute 路 Create Wisdom Team Chicago, IL 路 Jan 2022 - May 2023 Developed an ML-powered medical search platform that combined clinical data, content understanding, and semantic retrieval to improve access to relevant healthcare information. |
Academic ServiceReviewer, ACM Transactions on Knowledge Discovery from Data (TKDD), 2025 |
Teaching
Teaching Assistant, CompSci 516: Database Systems, Duke University, Spring 2025
|
Honors & Awards
Graduate School Fellowship, Duke University, 2023 - 2024
|
|
Last updated: July 2026. 漏 2026 Zhuofan Jia. Original template based on Jon Barron. |