Kehui Tan

Curriculum Vitae

Education

Peking University — Master’s student, 2025–2028 (expected)
Open Lab on Human Robot Interaction (PKU-HRI)
Advisor: Prof. Hong Liu

South China Normal University — Undergraduate study in Artificial Intelligence

Research interests

Multimodal AI, large language models, and embodied AI. Related experience includes visual geometry, trajectory prediction, and retrieval-augmented generation.

Selected research

Multi-View Depth Estimation

Combines geometric consensus-aware cost aggregation with geometrically anchored feature rectification to improve multi-view depth estimation.

Human Trajectory Prediction

Uses a diffusion-based preference score as a reward for trajectory prediction, together with rejection sampling.

Retrieval-Augmented Medical Question Answering

Improves medical document retrieval through embedding fine-tuning for retrieval-augmented query answering.

LLM Generation and Evaluation

Uses selected teaching examples and prompting to improve classroom dialogue completion and expertise transfer, with automatic and human evaluation.

Multi-Domain Retrieval

Explores mixture-of-experts modeling for retrieval across multiple domains.

Publications

Profiles

Google Scholar · ORCID · GitHub