Embedded Files
Multimodal understanding
We study how models connect language with what they see. Our work spans structured visual content such as tables and charts, as well as broader image and video understanding.
Table & Chart Understanding / Document VQA / Vision-Language Models / Visual Reasoning
Code intelligence
Understanding and generating code is at the core of this direction. We investigate how AI reasons about programs and bridges natural and programming languages.
Code Generation / Program Reasoning / Repository-level Understanding / Code Agents
Trustworthy AI
We explore the reliability, safety, and fairness of language and multimodal models, aiming for systems that are not only accurate but dependable across users and domains.
Robustness / Evaluation & Benchmarks / LLM-as-a-Judge / Safety & Alignment / Bias & Fairness
AI for education
We build AI systems that support how people actually learn, modeling learner knowledge from real submissions and designing tutors that adapt guidance to each student.
LLM Tutors / Programming Knowledge Tracing / Learner Modeling / Pedagogical Alignment
Recommender systems
Personalization lies at the heart of modern AI applications. We focus on systems that adapt to evolving user behavior, including conversational settings where needs surface through dialogue.
Sequential Recommendation / LLM-based Recommendation / Conversational Recommendation
Page updated
Google Sites
Report abuse