College AI · Computer Science and Technology
Chu Kochen Honors College · Artificial Intelligence (Turing Class)
arXiv2026
Presents an open benchmark and evaluation platform for Brain Foundation Models on electrical brain signals, organizing the field through a self-supervised learning taxonomy and a task-and-dataset view. The work integrates 15 representative BFMs with 18 public datasets to analyze how pretraining data, SSL strategy, and model architecture influence downstream generalization and transfer.
NeurIPS2025
Proposes a closed-loop full-lifecycle framework for embodied lifelong learning, arguing that real progress requires tightly coupling data acquisition, scene reconstruction and augmentation, shared multimodal representation learning, and sim-from-real evaluation. Arcadia shows that this end-to-end improvement loop yields consistent gains on navigation and manipulation benchmarks while transferring robustly to physical robots.
CVPR2025
Introduces a reasoning-augmented geo-localization framework built around three components: the GRE30K reasoning dataset, a multi-stage GRE model for progressively inferring scene attributes and local details, and GREval-Bench for evaluating localization from continent and country level down to city and street level. The method improves both accuracy and interpretability over prior VLM-based geo-localization approaches.
Developed an archaeology toolkit using image segmentation foundation models for document layout analysis and information extraction from archaeological reports.
An AI-powered project developed for the Douyin AI Creator competition, creating innovative solutions through generative AI technologies.
Constructed ReVQ-2k dataset (2,000 videos, 57,450 ratings) and developed a dual-stream no-reference video quality assessment model.
Only three recipients in the college, recognizing top-tier academic performance and overall excellence.
Awarded at the national level for innovation-oriented project development and interdisciplinary execution.
Received in consecutive years, reflecting consistent academic standing and sustained undergraduate performance.
Recognized for strong quantitative modeling, problem decomposition, and technical communication in an international contest setting.
Awarded for exceptional foundational performance in core computer science coursework during undergraduate study.
Earned provincial first prize, highlighting an early and strong mathematical foundation before university-level research training.
Teaching
Zhejiang University, College of Computer Science
Python Programming, Computer Systems I, Computer Organization
ZJU X-Lab(浙江大学启真交叉学科创新创业实验室)
Software development projects
Internship
Embodied AI · Vision-Language Navigation (VLN)
Building simulation environment frameworks for RL training and VLN