Welcome to my homepage! I am currently a research engineer at Beijing Institute for General Artificial Intelligence. My current research interests include: Embodied AI, Video Understanding and Robotics. Previously I worked on Bioinformatics and Graph Representation learning. In my leisure time, I am also intrested in Quantitative Trading.

Education

2016.9-2020.6, Bachelor in Computer Science, Peking University
2020.9-2023.6, Master in Intelligence Science and Technology, Peking University

Experiences

2022.7-2022.12, Research Intern, Microsoft Research Aisa.
We won the second place in the NeurIPS 2022 OGB-LSC challenge, which is a competition of link prediction on huge knowledge graphs. Our technical report can be found here.

Publications

Yue Fan, Xiaojian Ma, Rujie Wu, Yuntao Du, Jiaqi Li, Zhi Gao, Qing Li. VideoAgent: A Memory-augmented Multimodal Agent for Video Understanding. ECCV 2024. [pdf]
We build VideoAgent that understands long videos and achieves the performance close to Gemini 1.5 Pro. Details can be found here. We are now building real-time VideoAgent.

Jun Guo, Xiaojian Ma, Yue Fan, Huaping Liu, Qing Li. Semantic Gaussians: Open-Vocabulary Scene Understanding with 3D Gaussian Splatting. arxiv 2024. [pdf]
This work reconstructs the 3D scenes and perserves the semantic information using Gaussian splatting.

Tianxu Wang, Yue Fan, Xiuli Ma. Attention Based Models for Cell Type Classification on Single-Cell RNA-Seq Data. ECAI 2023. [pdf]
We build two kinds of attention-based model to predict the types of the cells. We found the attention weights to be a good interpretation for cell-type classification. This paper is awarded as the outstanding paper for AI in socal good.

Yue Fan, Xiuli Ma. Multi-vector embedding on networks with taxonomies. IJCAI 2022. [pdf]
We build a novel graph embedding method that adaptively learns multiple representation vectors for each node in a hyperbolic space.

Yue Fan, Xiuli Ma. Gene regulatory network inference using 3D convolutional neural network. AAAI 2021. [pdf]
A 3D convolutional neural network is built and serves a good gene regulatory network predictor.

Junshan Wang, Zhicong Lu, Guojia Song, Yue Fan, Lun Du, Wei Lin. Tag2vec: Learning tag representations in tag networks. WWW 2019. [pdf]
Tag2Vec learns the embedding of both the nodes and their tags in a hyperbolic space.