Xin Dong

Xin Dong

Cambridge, MA, May 2020

I am a research scientist at NVIDIA Research. I received Ph.D. in Computer Science from Harvard University in 2023, advised by H.T. Kung. I have general research interests in deep learning, with a focus on designing accurate, efficient and trustworthy systems for autonomous machines, LLM and GenAI.

Prior to Harvard, I was a research assistant at Nanyang Technological University and UC San Diego. I am a recipient of the Harvard James Miller Peirce Fellowship.

google scholar github twitter

Email: (firstname)d [at] nvidia.com

Research intern positions (2025) at NVIDIA available focused on foundation models. Please drop your resume to {firstname}d@nvidia.com if interested.

News

Experiences

NVIDIA
Research Scientist, 2024 - Present
Sony
Research Scientist, 2023 - 2024
Meta Reality Lab
Research Scientist Intern, 2021, 2022
NVIDIA
Research Scientist Intern, 2020
Tencent America
Research Scientist Intern, 2019

Publications

gpt-pruning
A Deeper Look at Depth Pruning of LLMs
ICML 2024 Workshop on Theoretical Foundations of Foundation Models (ICML Workshop 2024)
Shoaib Ahmed Siddiqui, Xin Dong, Greg Heinrich, Thomas Breuel, Jan Kautz, David Krueger, Pavlo Molchanov
gpt-pruning
The Cost of Down-Scaling Language Models: Fact Recall Deteriorates before In-Context Learning
International Conference on Learning Representations (ICLR 2024)
Tian Jin*, Nolan Clement*, Xin Dong*, Vaishnavh Nagarajan, Michael Carbin, Jonathan Ragan-Kelley, Gintare Karolina Dziugaite
(*: equal contribution)
gpt-pruning
Finding needles in a haystack: A Black-Box Approach to Invisible Watermark Detection
2024 European Conference on Computer Vision (ECCV 2024)
Minzhou Pan, Zhenting Wang, Xin Dong, Vikash Sehwag, Lingjuan Lyu, Xue Lin
gpt-pruning
A Simple Background Augmentation Method for Object Detection with Diffusion Mode
2024 European Conference on Computer Vision (ECCV 2024)
Yuhang Li, Xin Dong, Chen Chen, Weiming Zhuang, Lingjuan Lyu
gpt-pruning
Segment Every Out-of-Distribution Object
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2024)
Wenjie Zhao, Jia Li, Xin Dong, Yu Xiang, Yunhui Guo
fedicon
Is Heterogeneity Notorious? Taming Heterogeneity to Handle Test-Time Shift in Federated Learning
Conference on Neural Information Processing Systems (NeurIPS 2023)
Yue Tan, Chen Chen, Weiming Zhuang, Xin Dong, Lingjuan Lyu, Guodong Long
mTVR
Privacy Vulnerability of Split Computing to Data-Free Model Inversion Attacks
British Machine Vision Conference (BMVC 2022)
Xin Dong, Hongxu Yin, Jose M. Alvarez, Jan Kautz, Pavlo Molchanov, H.T. Kung
mTVR
SphereFed: Hyperspherical Federated Learning
European Conference on Computer Vision (ECCV 2022)
Xin Dong, Sai Qian Zhang, Ang Li, H.T. Kung
mTVR
SplitNets: Designing Neural Architectures for Efficient Distributed Computing on Head-Mounted Systems
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2022)
Xin Dong, Barbara De Salvo, Meng Li, Chiao Liu, Zhongnan Qu, H.T. Kung, Ziyun Li
VL-T5
Neural Mean Discrepancy for Efficient Out-of-Distribution Detection
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2022)
Xin Dong, Junfeng Guo, Ang Li, Wei-Te Ting, Cong Liu, H.T. Kung
VL-T5
DRESS: Dynamic REal-time Sparse Subnets
Efficient Deep Learning for Computer Vision (ECV) CVPR Workshop (ECV 2022)
Zhongnan Qu, Syed Shakib Sarwar, Xin Dong, Yuecheng Li, Ekin Sumbul, Barbara De Salvo
VL-T5
Training for Multi-resolution Inference Using Reusable Quantization Terms
The 26th ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS 2021)
Sai Qian Zhang, Bradley McDanel, H.T. Kung, Xin Dong
VL-T5
A free lunch from ANN: Towards efficient, accurate spiking neural networks calibration
International Conference on Machine Learning (ICML 2021)
Yuhang Li, Shikuang Deng, Xin Dong, Ruihao Gong, Shi Gu
VL-T5
MixMix: All You Need for Data-Free Compression Are Feature and Data Mixing
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2021)
Yuhang Li, Feng Zhu, Ruihao Gong, Mingzhu Shen, Xin Dong, Fengwei Yu, Shaoqing Lu, Shi Gu
MART
Additive Powers-of-Two Quantization: A Non-uniform Discretization for Neural Networks
International Conference on Learning Representations (ICLR 2020)
Yuhang Li*, Xin Dong*, Wei Wang
(*: equal contribution)
TVQA PLUS
RTN: Reparameterized Ternary Network
The Thirty-Fourth AAAI Conference on Artificial Intelligence (AAAI 2020)
Yuhang Li*, Xin Dong*, Sai Qian Zhang, HaoliBai, Yuanpeng Chen, Wei Wang
(*: equal contribution)
VLEP
exBERT: Extending Pre-trained Models with Domain-specific Vocabulary Under Constrained Training Resources
The 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP 2020)
Wen Tai, H.T. Kung, Xin Dong, Marcus Comiter, Chang-Fu Kuo
TVR
Efficient Bitwidth Search for Practical Mixed Precision Neural Network
arXiv 2020
Yuhang Li, Wei Wang, Haoli Bai, Ruihao Gong, Xin Dong, Fengwei Yu
TVR
Differentiable Dimension Search for Binary Neural Networks
1st Workshop on Neural Architecture Search at ICLR 2020 (ICLR 2020 Workshops))
Yuhang Li, Ruihao Gong, Fengwei Yu, Xin Dong, Xianglong Liu
TVQA
A Main/Subsidiary Network Framework for Simplifying Binary Neural Network
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2019)
Yinghao Xu*, Xin Dong*, Yudian Li, Hao Su
(*: equal contribution)
benn
Binary Ensemble Neural Network: More Bits per Network or More Networks per Bit?
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2019)
Shilin Zhu, Xin Dong, Hao Su
image classification
Full-stack Optimization for Accelerating CNNs with FPGA Validation
The ACM International Conference on Supercomputing (ICS 2019)
Bradley McDanel, Sai Qian Zhang, H.T. Kung, Xin Dong
image classification
Maestro: A Memory-on-Logic Architecture for Coordinated Parallel Use of Many Systolic Arrays
The 30th IEEE International Conference on Application-specific Systems, Architectures and Processors (ASAP 2019)
Bradley McDanel, Sai Qian Zhang, H.T. Kung, Xin Dong
image classification
Learning to Prune Deep Neural Networks via Layer-wise Optimal Brain Surgeon
Thirty-first Conference on Neural Information Processing Systems (NeurIPS 2017)
Xin Dong, Shangyu Chen, Sinno Jialin Pan

Academic Services