| Author: | Zhou, Huachi |
| Title: | Large language model-based recommender systems : integrating user-item features and interaction graphs into LLMs |
| Advisors: | Huang, Xiao (COMP) |
| Degree: | Ph.D. |
| Year: | 2026 |
| Department: | Department of Computing |
| Pages: | xiii, 104 pages : color illustrations |
| Language: | English |
| Abstract: | Large Language Models (LLMs) have recently been introduced into recommender systems to leverage their strong abilities in semantic understanding and reasoning beyond traditional models, raising the vision that the next generation of recommender systems could be LLM-based. However, simply plugging LLMs into recommendation pipelines, e.g., description-based prompting, raises three fundamental challenges: it is unclear whether LLMs truly learn the relationship between user–item features and interaction labels, how LLMs can directly exploit high-order structure in attributed graphs, and how to make such graph-integrated LLM recommenders efficient enough for real-world deployment. This thesis presents a systematic framework to address these challenges, advancing LLM-based recommendation by integrating user–item features into LLMs, extending this integration to attributed graphs, and ultimately enabling graph-based collaborative filtering (CF) on user–item interaction graphs with LLMs. First, we propose a self-monitoring framework that guides LLMs to align their predictions with informative user–item features. Next, we move to attributed graphs and introduce graph as a new language, enabling LLMs to capture high-order structure from new language corpus. Then we apply this idea to user–item interaction graphs and efficiently learn collaborative user/item embeddings from a sampled corpus for deployment. These three contributions form a coherent pipeline, demonstrating that with proper integration of user–item features and interaction graphs into LLMs, they can move beyond description-based prompting toward practical, scalable LLM-based recommender systems. |
| Rights: | All rights reserved |
| Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/14654