赵知临

副教授

联系邮箱: zhaozhlin@mail.sysu.edu.cn

联系地址: 计算机学院大楼A613

个人主页: https://lawliet-zzl.github.io/

教师简介

赵知临,研究员、副教授、博导、深圳河套学院双聘教授、国家级高层次青年人才

 Intelligence SciencE and systEm Lab (iSEE) 

研究领域

我们的研究主要聚焦于机器学习,特别关注模型在未知环境中的行为与泛化能力,追求大道至简的研究理念,注重理论与实践的结合,致力于在算法原理与现实应用之间架起桥梁。当前研究方向包括但不限于:

  • 机器学习基础:外分布泛化、鲁棒性优化、不确定性建模、时序分析等
  • 生成模型:图像生成、可控生成、图像编辑、生成对比学习等
  • 大语言模型:检索增强生成、幻觉检测与校正、跨语言与跨任务迁移
  • 具身智能与多智能体系统:虚实迁移与策略适应、语言驱动的具身操作、多智能体协同与决策等
  • 自由探索:如果你有自己的研究兴趣,也欢迎提出,我们可以一起学习、共同深入

招生与培养

欢迎对科研和技术感兴趣的博士生、硕士生和优秀本科生加入团队。相比一开始就掌握很多知识,我更看重好奇心、主动性,以及愿不愿意把想法真正做出来。科研中走弯路、实验没有达到预期都很正常,有新的想法也可以随时提出来,大家一起讨论、一起尝试。在培养上,我会根据每位同学的兴趣和未来规划,一起确定研究课题和阶段目标,并根据进展及时调整技术路线。能提前完成的科研任务会尽量往前安排,为之后的实习、求职或继续深造留出时间。团队目前主要关注大语言模型与多智能体、具身智能与机器人等方向。希望大家在完成论文和项目的同时,也能真正提升发现问题、动手实验、使用工具和写作表达的能力。实验室有比较充足的计算资源,也配备了多种机器人实验平台,可以支持大家把想法一步步做出来。

👉 点击 这里 查看我们团队的介绍。点击 这里 查看招生常见问题 Q&A,强烈建议在给我发邮件咨询之前,先完整阅读这份材料。

📢 目前有2027年入学的专业硕士(剩0)、学术硕士(剩0)、学术博士(剩1)、深圳河套学院联培博士名额(剩1)、医学交叉专项(剩1),如有意向,请随时与我联系。15622736176,微信同号。  

🙏 由于邮件较多,若您在一段时间内未收到回复,可能是邮件被遗漏,欢迎再次发送提醒。由此带来的不便,敬请谅解。

 

  

团队成员

如果你对加入我们或组内科研、学习情况感兴趣, 欢迎与我联系,也欢迎直接联系任何一位在组学生了解更多信息。

年级姓名类型毕业院校研究方向邮箱
2025张悦熠硕士南京理工大学跨形态的通用世界模型zhangyy696@mail2.sysu.edu.cn 
李姝萱硕士湖南大学性能预测与多词元预测lishx87@mail2.sysu.edu.cn 
王子峻硕士暨南大学生成模型与多模态分析wangzj87@mail2.sysu.edu.cn 
周朗博士x深圳河套中山大学(本硕)面向长文本理解的高效建模与分解推理zhoulang3@mail2.sysu.edu.cn 
李秋霜博士x医学交叉南昌大学 / 中山大学转录异构体分辨率的细胞状态表征liqsh25@mail2.sysu.edu.cn






2026范沐声硕士中山大学时序分析与天气预测fanmsh3@mail2.sysu.edu.cn
朱炜豪硕士华南理工大学检索增强生成与人类反馈强化学习zhuwh55@mail2.sysu.edu.cn
龚明硕士湖南大学长期时域机器人记忆gongm26@mail2.sysu.edu.cn
吴昱欣硕士南开大学大语言模型安全wuyx383@mail2.sysu.edu.cn
王明辉硕士北京交通大学大模型空间理解与天气预测wangmh89@mail2.sysu.edu.cn
何霖濮博士郑州大学 / 电子科技大学动作因果与反事实潜在世界模型help9@mail2.sysu.edu.cn
陈英健博士x深圳河套河南大学(本硕)/
维多利亚大学(本科双学位)/
东京大学与新加坡国立大学(硕士访学)
多智能体的过程验证与因果信用分配chenyj659@mail2.sysu.edu.cn






 

主持/参与研究课题

  • 国家高层次青年人才项目(2024-2027),主持
  • 国家自然科学面上项目(2025-2028),主持
  • 广东省重点研发(2026-2029),课题负责人
  • 中山大学青年科学家项目(2025-2028),主持

获奖及荣誉

  • 2018 广东省优秀硕士毕业生
  • 2016 中山大学优秀本科毕业生

教育背景

  • 2018.8-2022.8,悉尼科技大学,工程与信息技术学院,博士
  • 2016.9-2018.6,中山大学,数据科学与计算机学院,硕士
  • 2012.9-2016.6,中山大学,移动信息工程学院,学士

关于中山大学与悉尼,我还留下了一些照片和故事,分别收录在 《路与天色》 和 《Between City and Sea》 中。

工作经历

  • 2025.10至今,深圳河套学院,具身智能与计算机视觉中心,副教授
  • 2024.11至今,中山大学,计算机学院,副教授
  • 2023.5.9-2024.11,麦考瑞大学,科学与工程学院,博士后
  • 2022.8-2023.5,悉尼科技大学,工程与信息技术学院,博士后

主讲课程

欢迎同学们通过企业微信或邮件向我反馈课件中存在的错误、疏漏之处,以及对课程内容和展示方式的改进建议。非常感谢大家的支持与帮助。

代表性论著

Authorship Clarification: All the publications listed below on which I am the first author were primarily completed during my PhD and postdoctoral research. Professor Longbing Cao, who is listed as the second author, was my academic supervisor during this period. For student-led work, my students take the first-author spot, while I happily move to the corresponding-author position (*).

 

[TPAMI] Zhilin Zhao, Longbing Cao, Yixuan Zhang, Kun-Yu Lin, Wei-Shi Zheng. Distilling the Unknown to Unveil Certainty. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), vol. 47, no. 10, pp. 9232-9249, 2025. 

[TPAMI] Zhilin Zhao, Longbing Cao, Kun-Yu Lin. Supervision Adaptation Balancing In-Distribution Generalization and Out-of-Distribution Detection. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), vol. 45, no. 12, pp. 15743-15758, 2023.

[TPAMI] Zhilin Zhao, Longbing Cao, Kun-Yu Lin. Revealing the Distributional Vulnerability of Discriminators by Implicit Generators. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), vol. 45, no. 7, pp. 8888-8901, 2022.

[AIJ] Zhilin Zhao, Longbing Cao, Philip S. Yu. Out-of-distribution Detection by Regaining Lost Clues. Artificial Intelligence Journal (AIJ), vol. 339, pp. 104275,2024.

[TMLR] Zhilin Zhao, Longbing Cao. Dual Representation Learning for Out-of-Distribution Detection. Transactions on Machine Learning Research (TMLR), pp.1-21, 2023.

[MLJ] Zhilin Zhao, Longbing Cao. Weighting Non-IID Batches for Out-of-distribution Detection. Machine Learning Journal (MLJ), vol. 113, no. 10, pp. 7371-7391, 2024.

[TNNLS] Zhilin Zhao, Longbing Cao, Kun-Yu Lin. Out-of-Distribution Detection by Cross-Class Vicinity Distribution of In-Distribution Data. IEEE Transactions on Neural Networks and Learning Systems (TNNLS), vol. 35, no. 10, pp. 13777-13788, 2024.

[TNNLS] Zhilin Zhao, Longbing Cao, Chang-Dong Wang. Gray Learning from Non-iid Data with Out-of-distribution Samples. IEEE Transactions on Neural Networks and Learning Systems (TNNLS), vol. 36, no. 1, pp. 1396-1409, 2025.

[NeurIPS] Zhilin Zhao,Longbing Cao, Xuhui Fan, Wei-Shi Zheng. Revealing Distribution Discrepancy by Sampling Transfer in Unlabeled Data. Advances in Neural Information Processing Systems (NeurIPS), pp. 1-28, 2024.

[NeurIPS] Zhilin Zhao,Longbing Cao. R-divergence for Estimating Model-oriented Distribution Discrepancy. Advances in Neural Information Processing Systems (NeurIPS), pp. 1-19, 2023.

[AAAI] Zhilin Zhao, Longbing Cao, Yuanyu Wan. Mixture of Online and Offline Experts for Non-stationary Time Series. Association for the Advancement of Artificial Intelligence (AAAI), pp. 1-8, 2025.

[IJCAI]  Zhilin Zhao, Longbing Cao, Philip S. Yu. Out-of-distribution Detection by Regaining Lost Clues. International Joint Conference on Artificial Intelligence (IJCAI),2025.

[IJCAI] Zhilin Zhao, Longbing Cao, Philip S. Yu. Deep Non-IID Learning. International Joint Conference on Artificial Intelligence (IJCAI), 2024 (Tutorial).

[CVPR] Shuxuan Li (M.S. student), Zhilin Zhao*, Quyu Kong, Wei-Shi Zheng. Bridging Domain Expertise and Generalization for Performance Estimation, IEEE/CVF Conference on Computer Vision & Pattern Recognition (CVPR), 2026.

[EMNLP] Lang Zhou (Ph.D. student), Shuxuan Li, Zhuohao Li, Shi Liu, Wei-Shi Zheng, Zhilin Zhao*. UT-ACA: Uncertainty-Triggered Adaptive Context Allocation for Long-Context Inference, Empirical Methods in Natural Language Processing (EMNLP), Main Conference, 2026.

 

A small note for careful readers: You may have noticed that my first paper in a top journal appeared only after I finished my PhD. Apparently, some of my papers took longer to graduate than I did. I was never a particularly fast researcher. I spent much of my PhD learning how research actually works, strengthening my theoretical foundations, and getting things wrong more times than I can count. Instead of writing one paper at a time like a sensible person, I tried to build a larger research framework first and turn it into several connected papers later. Efficient? Not really. Recommended? Probably not. Fortunately, my supervisor was patient, supportive, and kept encouraging me to pursue ideas that I believed were worth exploring. Many of those papers therefore appeared only after I graduated. So when I tell students that failure is normal in research, I really mean it. I am just an ordinary researcher who has been wrong many times, occasionally right, and stubborn enough to keep going.

“In the midst of winter, I found there was, within me, an invincible summer.”  -- Albert Camus