Shanghao (Cara) Li

About

I am a computer science Ph.D. candidate at the University of Illinois Chicago (UIC), fortunate to be advised by Prof. Philip S. Yu in his Big Data and Social Computing (BDSC) Lab.

Generally, my research lies at the intersection of Reinforcement Learning, Large Language Models(LLMs), and Human–Agent Interaction(HAI). More specifically, I develop methods to make LLMs more reliable, aligned, and effective collaborators in knowledge-intensive and multi-turn tasks.

Recently, I have been working on GraphRAG, creating metrics that analyze LLM attention patterns and semantic drift, and building efficient detectors for identifying reasoning drift from retrieved subgraphs. I am also developing TactfulLLM, a framework that trains LLMs to regulate their initiative by learning when to ask clarifying questions and when to act, using preference learning and multi-objective rewards.

News

• 2026/09-2026/12: PhD Machine Learning Intern at Pinterest Advanced Technologies Group (ATG - Visual, Multimodal, and Recommender Systems)
• 2026/10: One paper was accepted by NeurIPS 2026.
• 2026/07/05: Presented an accepted paper at the ACL conference in San Diego, California.
• 2026/05-2026/08: PhD Intern at Google
• 2026/04: One paper was accepted by ACL 2026 main track
• 2025/08/10: Invited to serve as Program Committee (PC) member for the ACM International Conference on Intelligent User Interfaces (IUI) 2026.
• 2024/05/11: Attended CHI 2024 in Hawaii and presented an accepted paper.
• 2024/03/01: One paper was accepted by ACM CHI 2024.
• 2023/08/18: Started my Ph.D. at the University of Illinois Chicago ~

Selected Publications

Why LLMs Hallucinate on Structured Knowledge: A Mechanistic Analysis of Reasoning over Linearized Representations
Shanghao Li, Jinda Han, Yibo Wang, Yuanjie Zhu, Zihe Song, Langzhou He, Kenan Kamel A Alghythee, Philip S. Yu
ACL 2026 Main ACL Anthology

Detecting Hallucinations in Graph Retrieval Augmented Generation via Attention Patterns and Semantic Alignment
Shanghao Li, Jinda Han, Yibo Wang, Yuanjie Zhu, Zihe Song, Langzhou He, Kenan Kamel A Alghythee, Philip S. Yu
arXiv 2025 [arXiv]

Towards Understanding Group Collaboration Patterns Around Mobile Augmented-Reality Interfaces for Geospatial Data Visualizations
Shanghao Li, Taylor Lane, Nuoran Chen, Alicia Hernandez, Vinayak Kabra, Karthik Singh, Stefany Sit and Nikita Soni
CHI 2024 [PDF] [ACM] [VIDEO] [POSTER]

Sentiment analysis and topic modeling regarding online classes on the Reddit Platform: educators versus learners
Shanghao Li, Zerong Xie, Dickson KW Chiu, Kevin KW Ho
Applied Sciences 2023 [PDF] [LINK]

Fun Fact

My cat creamsicle 🐱