Undergraduate Research Spotlight: Trusted AI in a Connected World: Evaluating Robustness of Graph Large Language Models

Join our PhD and Undergraduate Students Research Spotlight Series. This series provides a platform for students to share their work and connect with the academic community.

 

Graphs, especially text-attributed ones like amazon co-purchased product networks and citation networks, have become a vital part of our modern lives. Consequently, machine learning techniques that exploit these graph structures have surged in development. With the emergence of Large Language Models (LLMs), a novel paradigm—Graph Large Language Models (GraphLLMs)—has been introduced to tap into LLMs’ potential for interpreting graph-structured data. However, the robustness of these GraphLLMs remains relatively unexplored, posing challenges in evaluation as they integrate both textual and structural information. In this talk, we will dive into the challenges of assessing GraphLLM robustness and how to ensure these models perform reliably.

 

Xinyue Sheng is a junior student majoring in Data Science and Mathematics at New York University Shanghai. Under the supervision of Assistant Professor Qiaoyu Tan, her research interests lie in GraphLLMs, trustworthy AI, and multimodal generative AI for the relational world.