Computer Science Department , PhD Dissertation Defense, Kai Zhang "Memory-enhanced LLM Adaptations for Human-Centered Intelligence"

Kai Zhang 

PhD Candidate 

WPI – Computer Science Department 

Thursday, January 29, 2026

Time: 10:00 a.m. – 11:00 a.m. 

Zoom: https://wpi.zoom.us/j/93289149364

Committee members :

Prof. Xiaozhong Liu, WPI Computer Science/Data Science Department

Prof. Roee Shraga, WPI Computer Science/Data Science Department

Prof. Frank Zou, WPI Mathematical Science Department

Prof. Yejin Kim, UT Health, Department of Health Data Science and Artificial Intelligence

Abstract:

Large Language Models (LLMs) have demonstrated impressive capabilities in natural language understanding and generation. However, their ability to personalize responses and maintain contextual coherence across interactions remains limited. This proposal explores a memory-enhanced adaptation for LLMs, aiming to improve both the quality and personalization of responses by leveraging user-specific and community-guided memories. Our framework consists of three research questions (RQs): 

RQ1. Single-user memory, which retrieves and integrates long-term and short-term user-specific knowledge for tailored responses; 

RQ2. Community-guided memory, which augments user memory with shared knowledge from a relevant community to enhance response quality, especially in unseen scenarios; 

RQ3. In-memory knowledge reasoning, which structures memory-based information to support context-aware, topic-based reasoning. 

This study seeks to develop efficient retrieval mechanisms, privacy-preserving memory sharing, and structured reasoning to enable more intelligent and user-centric LLM interactions.