We advance personalized intelligence by uniting recommendation, multimodal learning, foundation models, and AI agents. This integrated approach transforms user signals into trusted predictions, generative experiences, and adaptive interactions.
Research Areas
Our research connects recommendation, multimodal intelligence, foundation models, and AI agents to create systems that understand individual needs and deliver adaptive, trustworthy experiences.
We develop personalized recommender systems that learn individual preferences from behavioral sequences, contextual signals, and relational graphs, then translate them into relevant rankings and adaptive experiences across changing user needs.
We combine language, images, reviews, and behavioral signals to construct richer representations of users and content. This multimodal perspective supports accurate recommendation, review understanding, and reliable detection of deceptive information.
We adapt large language models and foundation models to understand preferences and generate personalized content, explanations, and recommendations. Our work connects efficient model adaptation with reliable and contextually relevant generation.
We design personalized AI agents that retrieve relevant knowledge, reason about user goals, and communicate through natural conversation. These systems coordinate memory, context, and adaptive decision making to provide useful assistance.
Research Outcomes
Representative studies that demonstrate how our research advances language models, personalized recommendation, multimodal intelligence, and trustworthy information analysis.
Presents an efficient instruction tuned LLM framework for review helpfulness prediction under limited supervision.
Li, X., Li, Q., & Kim, J. (2026). Instruction-tuned large language models for review helpfulness prediction: An efficient fine-tuning framework for e-commerce review understanding. Expert Systems, 43, e70208.
View Paper
Examines how bot assisted fake engagement redirects public attention and shapes digital information consumption.
Lee, S., Shin, D., Kwon, K. H., Han, S. P., & Lee, S. K. (2024). Disinformation spillover: Uncovering the ripple effect of bot-assisted fake social engagement on public attention. MIS Quarterly, 48(3), 847-872.
View Paper
Improves review based recommendation by extracting aspect terms and filtering irrelevant contextual noise.
Lim, H., Li, X., Park, S., Li, Q., & Kim, J. (2026). Reducing contextual noise in review-based recommendation via aspect term extraction and attention modeling. Information Sciences, 735, 123078.
View Paper
Combines textual, visual, and review quality cues through attention and multimodal gating for helpfulness prediction.
Lim, H., Park, S., Li, Q., Li, X., & Kim, J. (2026). What makes a review helpful? A multimodal prediction model in e-commerce. Electronic Commerce Research and Applications, 76, 101586.
View PaperResearch Story
Our research transforms user, content, and contextual signals into relevant recommendations, generative experiences, and adaptive interactions through a connected AI process.
User Modeling · Multimodal Understanding · Context Awareness
Preference Learning · Personalized Ranking · Relevance Prediction
Foundation Models · Generative Recommendation · RAG
AI Agents · Conversational AI · Adaptive Interaction
Research Frontiers
We explore emerging AI technologies that extend our research in recommendation, multimodal intelligence, foundation models, agents, and review analysis to create more adaptive and trustworthy personalized AI.
Core Technologies
Our technology stack connects advanced AI models with practical development frameworks, enabling us to design, train, evaluate, and deploy personalized intelligence from research prototypes to real services.
PyTorch
TensorFlow
Keras
scikit-learn
Hugging Face
LangChain
OpenCV
FastAPI
MLflow
vLLM