苏骏炜｜中国科学技术大学｜学者主页

期刊论文

2026

Reward-Aligned Reweighting for On-Policy Distillation

Hao Xu, Junwei Su, Lan-Song Diao, Wen-Chao Zhou, Chuan Wu

2026

A Closed-Loop Non-Asymptotic Convergence Analysis of PPO with Learned Critics and Clipping

Junwei Su, Meng-Fan Liu, Yan-Yong Zhang, Chuan Wu

2026

ReFlow: Replay-Driven Decoupling for Efficient RL Post-Training

Yu Tian, Junwei Su, Hao Xu, Lan-Song Diao, Zheng-Ping Qian, Wei-Bo Zhao, Wen-Chao Zhou, Rui-Ting Zhou, Chuan Wu

2026

BG-HGNN: Toward Efficient Learning for Complex Heterogeneous Graphs

Junwei Su, Lingjun Mao, Da Zheng, Chuan Wu

Trans. Mach. Learn. Res.

2026

GAC: Stabilizing Asynchronous RL Training for LLMs via Gradient Alignment Control

Hao Xu, Junwei Su, Yu Tian, Lan-Song Diao, Zhe Qian, Chuan Wu

arXiv.org

DOI

2026

HetAuto: Cross-Cluster Auto-Parallelism for Heterogeneous Distributed Training

Guicheng Qi, Junwei Su, Liqiang Yang, Tao Li, Tingwen Xie, Yerui Sun, Yu-Chen Xie, Chuan Wu

European Conference on Computer Systems

DOI

2025

On the Interplay between Graph Structure and Learning Algorithms in Graph Neural Networks

Junwei Su, Chuan Wu

International Conference on Machine Learning

DOI

2025

A Non-Asymptotic Convergent Analysis for Scored-Based Graph Generative Model via a System of Stochastic Differential Equations

Junwei Su, Chuan Wu

International Conference on Machine Learning

DOI

2025

SBGD: Improving Graph Diffusion Generative Model via Stochastic Block Diffusion

Junwei Su, Shan Wu

International Conference on Machine Learning

DOI

2024

Towards Robust Graph Incremental Learning on Evolving Graphs

Junwei Su, Difan Zou, Zijun Zhang, Chuan Wu

International Conference on Machine Learning

DOI

2024

Temporal-Aware Evaluation and Learning for Temporal Graph Neural Networks

Junwei Su, Shan Wu

AAAI Conference on Artificial Intelligence

DOI

2024

Improving Implicit Regularization of SGD with Preconditioning for Least Square Problems

Junwei Su, Difan Zou, Chuan Wu

SIAM Journal on Mathematics of Data Science

DOI

2024

HEGformer: Representation Learning on Heterogeneous Graph Neural Network with Efficient Transformer

Tianzuo Qin, Junwei Su, Chuan Wu

2024 IEEE International Conference on Knowledge Graph (ICKG)

DOI

2024

Heta: Distributed Training of Heterogeneous Graph Neural Networks

Yu-Chen Zhong, Junwei Su, Chuan Wu, Minjie Wang

Proceedings of the VLDB Endowment

DOI

2024

On the Topology Awareness and Generalization Performance of Graph Neural Networks

Junwei Su, Chuan Wu

European Conference on Computer Vision

DOI

2024

MTRGL: Effective Temporal Correlation Discerning Through Multi-Modal Temporal Relational Graph Learning

Junwei Su, Shan Wu, Jinhui Li

IEEE International Conference on Acoustics, Speech, and Signal Processing

DOI

2024

PRES: Toward Scalable Memory-Based Dynamic Graph Neural Networks

Junwei Su, Difan Zou, Chuan Wu

International Conference on Learning Representations

DOI

2024

MSPipe: Efficient Temporal GNN Training via Staleness-Aware Pipeline

Guang-Ming Sheng, Junwei Su, Chao Huang, Chuan Wu

Knowledge Discovery and Data Mining

DOI

2023

CDMPP: A Device-Model Agnostic Framework for Latency Prediction of Tensor Programs

Hanpeng Hu, Junwei Su, Juntao Zhao, Yanghua Peng, Yi-Bo Zhu, Haibin Lin, Chuan Wu

European Conference on Computer Systems

DOI

2023

On the Limitation and Experience Replay for GNNs in Continual Learning

Junwei Su, Difan Zou, Chuan Wu

CoLLAs

2022

ABC: Aggregation before Communication, a Communication Reduction Framework for Distributed Graph Neural Network Training and Effective Partition

Junwei Su

arXiv.org

DOI

预印本

2026

Structuring Relations Among Learning Paradigms via Protocol--Objective--Resource Reductions

Junwei Su, Changjie Wang, Dongyang Chang

arXiv (Cornell University)

DOI

苏骏炜

精选作品

期刊论文

Reward-Aligned Reweighting for On-Policy Distillation

A Closed-Loop Non-Asymptotic Convergence Analysis of PPO with Learned Critics and Clipping

ReFlow: Replay-Driven Decoupling for Efficient RL Post-Training

BG-HGNN: Toward Efficient Learning for Complex Heterogeneous Graphs

GAC: Stabilizing Asynchronous RL Training for LLMs via Gradient Alignment Control

HetAuto: Cross-Cluster Auto-Parallelism for Heterogeneous Distributed Training

On the Interplay between Graph Structure and Learning Algorithms in Graph Neural Networks

A Non-Asymptotic Convergent Analysis for Scored-Based Graph Generative Model via a System of Stochastic Differential Equations

SBGD: Improving Graph Diffusion Generative Model via Stochastic Block Diffusion

Towards Robust Graph Incremental Learning on Evolving Graphs

Temporal-Aware Evaluation and Learning for Temporal Graph Neural Networks

Improving Implicit Regularization of SGD with Preconditioning for Least Square Problems

HEGformer: Representation Learning on Heterogeneous Graph Neural Network with Efficient Transformer

Heta: Distributed Training of Heterogeneous Graph Neural Networks

On the Topology Awareness and Generalization Performance of Graph Neural Networks

MTRGL: Effective Temporal Correlation Discerning Through Multi-Modal Temporal Relational Graph Learning

PRES: Toward Scalable Memory-Based Dynamic Graph Neural Networks

MSPipe: Efficient Temporal GNN Training via Staleness-Aware Pipeline

CDMPP: A Device-Model Agnostic Framework for Latency Prediction of Tensor Programs

On the Limitation and Experience Replay for GNNs in Continual Learning

ABC: Aggregation before Communication, a Communication Reduction Framework for Distributed Graph Neural Network Training and Effective Partition

预印本

Structuring Relations Among Learning Paradigms via Protocol--Objective--Resource Reductions

研究方向