This page lists my selected manuscripts and preprints. Some works are under review or under preparation, and are therefore not formal peer-reviewed publications.
This work investigates bidirectional interaction between scene parsing and geometric vision tasks, including semantic segmentation, instance segmentation, stereo matching, and optical flow.
This work empirically studies the reliability of large language models in recommending software libraries and supporting code migration tasks.
This work studies cinematic language in text-to-video generation, focusing on shot framing, angles, and camera movements. It introduces a cinematic language dataset, CameraCLIP for alignment evaluation, and CLIPLoRA for dynamic cinematic style composition.