About Me

你好，我是赵嘉乐，目前就读于广东工业大学。我的研究兴趣主要围绕机器学习及其在安全与可靠性方面的应用，尤其关注模型在复杂真实环境中的行为与边界。

从大一开始，我有幸在朱鉴老师的指导下接触 3D Vision 相关研究工作，这段经历让我初步建立了对数据表示与建模问题的直观认识，也让我开始意识到现实数据与理想假设之间的差距。随后，我在 Zhao Song 老师的指导下开展了 Machine Learning Theory 与 Differential Privacy 方面的研究，从更加理论的角度思考模型的泛化能力与隐私约束。基于这些经验，我近期将研究重点进一步延伸至 AI Safety 与 AIGC Detection 方向，并在秦景辉老师的指导下关注对抗环境下模型的鲁棒性与安全性问题。这些经历逐步形成了一条从表示学习、到理论分析、再到安全与可信性的研究路径。

我性格踏实，做事专注，也在不断反思与改进自己的节奏与方法，希望能够在长期科研中保持稳定的投入与成长。目前我正在积极准备 2026 Fall 保研，期待能够加入合适的研究团队，在老师的指导下继续深入相关方向。

News✨

  • 26/08 One paper has been accepted by Neural Networks (CCF-B)
  • 26/07 One paper has been accepted by ACM MM 2026 (CCF-A)
  • 26/05 Google Scholar 100 citations 🎉
  • 25/12 One paper has been accepted by The Visual Computer (CCF-C)
  • 25/10 One paper has been accepted by BigData 2025 (CCF-C)
  • 25/08 One paper has been accepted by ICDM 2025 (CCF-B) (13.5%接受率)
  • 25/08 One paper has been accepted by EMNLP 2025 Findings (CCF-B)

Publications (* indicates equal contribution, alphabetical order)

ISPCloak: Weaponizing ISP for Optimization-Free Physical Camouflage against Deepfake Detectors

Jiale Zhao, Jiajun Wan, Lei Tang, Ye Qin, Kebing Jin, Jinghui Qin

ACM Multimedia 2026

[ Paper]
Fast Sampling for Privacy-Preserving Lazy Multiplicative Weight Update

Xiaoyu Li*, Zhao Song*, Jiale Zhao*

ICDM 2025

[ Paper]
Conv-Basis: A New Paradigm for Efficient Attention Inference and Gradient Computation in Transformers

Yingyu Liang*, Heshan Liu*, Zhenmei Shi*, Zhao Song*, Zhuoyan Xu*, Jiale Zhao*, Zhen Zhuang*

EMNLP 2025 Finding

[ Paper]
A Theoretical Framework Bridging Attention and SVM Optimization Dynamics and Sparsity

Zhihang Li*, Zhao Song*, Jiale Zhao*

BigData 2025

[ Paper]
DWGLT: a deformable window-based global–local transformer network for efficient image super-resolution

Jian Zhu, Simin Zheng, Peiyao Chen, Jiale Zhao, Zhaohui Liao, Rongqiu Wang, Guang Feng

The Visual Computer

[ Paper]
T2vtextbench: A human evaluation benchmark for textual control in video generation models

Xuyang Guo*, Jiayan Huo*, Zhenmei Shi*, Zhao Song*, Jiahao Zhang*, Jiale Zhao*

Preprint

[ Paper]
T2vphysbench: A first-principles benchmark for physical consistency in text-to-video generation

Xuyang Guo*, Jiayan Huo*, Zhenmei Shi*, Zhao Song*, Jiahao Zhang*, Jiale Zhao*

Preprint

[ Paper]
Fundamental Limits of Crystalline Equivariant Graph Neural Networks: A Circuit Complexity Perspective

Yang Cao*, Zhao Song*, Jiahao Zhang*, Jiale Zhao*

Preprint

[ Paper]
Evaluating Frontier LLMs on PhD-Level Mathematical Reasoning: A Benchmark on a Textbook in Theoretical Computer Science about Randomized Algorithms

Yang Cao*, Yubin Chen*, Xuyang Guo*, Zhao Song*, Song Yue*, Jiahao Zhang*, Jiale Zhao*

Preprint

[ Paper]

Honors & Awards

  • 2024华为昇腾AI创新大赛国赛优秀奖
  • 第十一届“挑战杯”学术科技作品竞赛校赛二等奖
  • 2024高教社杯全国大学生数学建模竞赛广东省三等奖
  • 2024华为昇腾AI创新大赛广州赛区金奖
  • 2023-2024年校级一等奖学金
  • 2024蓝桥杯C++赛道广东省一等奖