Publications

My current research interests:

• Agent Framework. [P5]
• Zeroth-Order Optimization. [P3, P7]
• Learning to Optimize. [P1, C9, C8, C7, C6, C5, W2, W1]
• AI for Computer Network Management & Scheduling. [P8, C9, C8, C7, C5, W2, W1]
• Optimization for Deep Learning. [C9, C8, P3, P7]
• Inference Acceleration for LLM. [P2]

My past research topics:

• Communication Efficient Federated Learning. [C4]
• Graph Neural Networks. [C5, W2, W1, C3, C1]
• Time Series Prediction. [C3, C1]

Preprints

[P8]. Siyong Huang, Qingyu Song*, Lizhao You, Qiang Su, Lu Tang, Wanjian Feng, Fei Yuan, Qiao Xiang*, Jiwu Shu, REACT: Toward Real-Time, End-to-End, Adaptive Cross-layer Restoration for IP-Over-Optical Networks.

[P7]. Wei Lin, Yining Jiang, Qingyu Song, Qiao Xiang, Hong Xu, AGZO: Activation-Guided Zeroth-Order Optimization for LLM Fine-Tuning. arXiv preprint arXiv:2601.17261 (2026)

[P6]. Mingyuan Song, Huan Shen, Jinghui Jiang, Qiang Su*, Qingyu Song, Lu Tang, Wanjian Feng, Fei Yuan, Qiao Xiang*, Jiwu Shu. Argo: An efficient verification framework for distributed in-network computing. arXiv preprint arXiv:2511.08189

[P5]. Yining Jiang, Yunxin Xu, Wenyun Xu, Yufan Zhu, Tangtang He, Haiying Huang, Letian Zhu, Qingyu Song*, Qiang Su, Lizhao You, Lu Tang, Wanjin Feng, Yuchao Zhang, Linghe Kong, Qiao Xiang*, Jiwu Shu. Leveraging Large Language Models for Automated Reproduction of Networking Research Results. arXiv preprint arXiv:2509.21074 (2026)

[P4]. Yao Wang, Kexin Yu, Wenyun Xu, Kaiqiang Hu, Ziyi Wang, Lizhao You*, Qiang Su, Dong Guo, Haizhou Du, Wanjian Feng, Qingyu Song, Linghe Kong, Qiao Xiang*, Jiwu Shu, Janus: Leveraging Incremental Computation for Efficient DNS Verification. arXiv preprint arXiv:2511.02559 (2025)

[P3]. Wei Lin, Qingyu Song, and Hong Xu. The Multi-Query Paradox in Zeroth-Order Optimization. arXiv preprint arXiv:2509.15552 (2025).

[P2]. Qingyu Song, Peiyu Liao, Wenqian Zhao, Yiwen Wang, Shoubo Hu, Hui-Ling Zhen, Ning Jiang, and Mingxuan Yuan. “Harnessing On-Device Large Language Model: Empirical Results and Implications for AI PC.” arXiv preprint arXiv:2505.15030 (2025). [Code]

[P1]. Wei Lin, Qingyu Song*, and Hong Xu. “Adaptive Coordinate-Wise Step Sizes for Quasi-Newton Methods: A Learning-to-Optimize Approach.” arXiv preprint arXiv:2412.00059 (2024).

Conference Proceedings (* Corresponding Author)

[C9]. Qiang Su, Yining Jiang, Siyong Huang, Qingyu Song*, Qiao Xiang*, Xue Liu, and Jiwu Shu. Toward Scalable and High-Performance GNN-Based Traffic Engineering with Free Path Selection. In IEEE ICPADS 2025. (CCF C)

[C8]. Qingyu Song, Wei Lin, and Hong Xu. Learning Provably Improves the Convergence of Gradient Descent. In NeurIPS 2025. [Paper] [Code] (CCF A)

[C7]. Siyong Huang, Qingyu Song, Kexin Yu, Zhaoning Wang, Zhizhen Zhong, Qiao Xiang, and Jiwu Shu. Toward Scalable Learning-Based Optical Restoration. In ACM APNet 2025. (CCF C)

[C6]. Qingyu Song, Wei Lin, Juncheng Wang, Hong Xu. Towards Robust Learning to Optimize with Theoretical Guarantees. In IEEE/CVF CVPR, 2024. [Paper] [Code] (CCF A)

[C5]. Qingyu Song, Juncheng Wang, Jingzong Li, Guocheng Liu, Hong Xu. A Learning-only Method for Multi-Cell Multi-User MIMO Sum Rate Maximization. In IEEE INFOCOM, 2024. (CCF A)

[C4]. Yu Zhang, Wei Lin, Sisi Chen, Qingyu Song, Jiaxun Lu, Yunfeng Shao, Bei Yu, Hong Xu. Fed2Com: Towards Efficient Compression in Federated Learning. In IEEE ICNC, 2024.

[C3]. Qingyu Song, RuiBo Ming, Jianming Hu, Haoyi Niu, Mingyang Gao. Graph Attention Convolutional Network: Spatiotemporal Modeling for Urban Traffic Prediction. In IEEE ITSC, 2020.

[C2]. Jinhua Chen, Qingyu Song, Can Zhao, Zhiheng Li. Graph Database and Relational Database Performance Comparison on a Transportation Network. In ICACDS, 2020.

[C1]. Qingyu Song, Jianming Hu, Ruobing Zhang, Zuo Zhang. An Urban Topological Map Generation Method for Traffic Flow Prediction Based on Road Segment Clustering with Floating Vehicle Trajectory Dataset. In COTA CICTP, 2019.

Workshops

[W2]. Qingyu Song, Guocheng Liu, Hong Xu. Learning to Optimize Non-Convex Sum-Rate Maximization Problems. In ICML 2023, 1st Workshop on Synergy of Scientific and Machine Learning Modeling. (CCF A)

[W1]. Qingyu Song, Guocheng Liu, Hong Xu. Towards a Learning-Only Approach for Non-Convex Sum Rate Maximization. In ACM SigMetrics 2023, 1st Workshop on Learning-augmented Algorithms: Theory and Applications. (CCF B)

You can also find my articles on my Google Scholar profile.