projects

Web Agent Evaluation & Benchmark

Process-level evaluation of web agents through semantic state tracking, enabling fine-grained diagnosis of exploration, execution, and decision-making failures.

Project page | Paper

3D Preference-Aligned Quality Assessment

Preference-aligned 3D quality assessment for scalable, human-aligned evaluation and curation of 3D assets.

Project page | Paper

Enhancing Text-to-3D Generation with PBR Materials

Reproduced baseline 3D generative models and prepared an Objaverse-scale 3D data pipeline including mesh processing, multi-view rendering, and PBR maps.

Developing a Sustainable Collaborative Multi-modal Continual Learning Framework

Built video data pipelines for UCF101 and HMDB51 and reproduced baseline continual learning methods for online continual learning research.

Synthetic Voice Detection

Proposed strategies for detecting AI-generated voice and placed third at the 2nd YAI-Con.