张林峰
助理教授
上海交通大学 人工智能学院
办公室：上海市徐汇区华山路1954号 人工智能学院 321-1
zhanglinfeng@sjtu.edu.cn

Unsplashed background img 1
-->


关于我

我当前在上海交通大学人工智能学院担任助理教授，领导 EPIC (Efficient and Precision Intelligent Computing) 实验室，有硕士与博士学生指导资格。 此前，我在2024年6月于清华大学交叉信息学院获得博士学位，导师为马恺声副教授。在读博期间，我获得了微软学者荣誉称号（亚太地区年度共十二名），北京市优秀毕业生，清华大学优秀博士论文，清华大学启航奖金奖，清华大学蒋南翔奖学金。在博士期间，我已发表高水平学术论文二十余篇，其中第一作者十三篇，论文总计被引用超过2200次（截至2024年11月）。本人研究成果已在北极雄芯、华为、交叉信息核心技术研究院等公司中得到使用。在2024年7月后，我以助理教授的身份加入上海交通大学人工智能学院。

当前，实验室正在招收博后、本科或研究生的研究助理以及2026/2027级的学生。如果你有兴趣，请查看我们的招生贴。

研究方向

一、轻量高效的语言/多模态大模型： 当前的生成式大模型有着百亿量级的参数，导致了极高的训练和推理成本，起了诸多问题。例如，OpenAI曾因无法承担计算成本限制其用户为ChatGPT4付费。通过研究面向生成式大模型的压缩加速方法，我们可以降低大模型的部署成本，使得大模型更好地在现实世界中得到使用。同时，如何使得小模型能具备大模型相同的表征能力也是人工智能的基础核心问题之一。

二、轻量高效的AIGC模型： 以Stable Diffusion和Sora为代表的文生图、文生视频模型已经掀起了AIGC（人工智能内容生成）的浪潮。然而，高分辨率的图像和长视频的计算成本往往极高，导致其难以真正投入到产业应用中。为解决该问题，我们致力于实现实现高效的视觉生成模型，推动AIGC产业落地。

三、数据高效的人工智能： 当前的人工智能模型需要在极大量的数据基础上进行训练，这严重提高了大模型的训练成本。如何更加高效地利用数据，更加科学地对数据进行清洗和合成，以及利用合成数据进一步提升生成式模型，通往数据高效的人工智能。

学术服务

在以下学术会议和期刊担任审稿人： NeurIPS, ICML, ICLR, CVPR, ECCV, ICCV, AAAI, IJCAI, AISTATS, IEEE TPAMI, IEEE TCSVT, IEEE TIP, IEEE TMI, IJCV, Pattern Recognition, TACO, Scientific Reports and others.

在以下学术会议和期刊担任领域主席或客座编辑： IJCNN2025, Big Data and Cognitive Computing, ACL2025.



近期新闻
  •   • 3 / 2025:   实验室十篇文章投稿于ICCV2025，祝同学们好运！
  •   • 3 / 2025:   实验室四篇文章被CVPR2025，其中一篇收获满分，恭喜！
  •   • 2 / 2025:   实验室三篇文章投稿于ACL2025，祝同学们好运！
    •   • 1 / 2025:   实验室三篇文章被ICLR2025录用，一篇被选为Oral，恭喜！
      •   • 12 / 2024:   实验室一篇文章被AAAI2025录用，恭喜！
      •   • 12 / 2024:   张林峰的相关信息由澎湃新闻报导，点击量过亿，登录微博热搜第一，知乎热搜第一，由人民日报等媒体报导。
      •   • 11 / 2024:   实验室七篇论文投稿于CVPR2025，祝同学们好运。
      •   • 10 / 2024:   实验室一篇文章被NeurIPS2024录用，五篇论文投稿于ICLR2025，祝同学们好运。
      •   • 09 / 2024:   实验室2025届研究生招生工作结束，欢迎温子辰，李学林，闫泽轩同学加入EPIC Lab。
      •   • 09 / 2024:   实验室两篇论文投稿于AAAI2025。
      •   • 08 / 2024:   在胡旭明教授的邀请下，张林峰赴香港科技大学（广州）兼任访问助理教授。
      •   • 06 / 2024:   张林峰在清华大学交叉信息研究院获得工科博士学位，同时获北京市优秀毕业生，清华大学优秀博士论文，清华大学启航奖金奖，交叉信息研究院优秀毕业生，并作为学院代表在参与清华大学毕业生座谈会，在毕业典礼上发言。


邀请报告

2024.12， 东北大学，沈阳，报告题目：免训练的扩散模型推理加速技术。

2024.12， 华为车智能汽车BU AI智能青年论坛，上海，报告题目：词元视角下的生成式模型压缩。

2024.12， 上海财经大学，上海，报告题目：词元视角下的生成式模型压缩。

2024.11， 中国农业大学，上海，报告题目：基于扩散模型的AIGC模型加速推理加速。

2024.4， 华为计算产品线青年论坛，杭州，报告题目：基于知识蒸馏的模型压缩。

主要成员
Linfeng Zhang
Research Interest: Efficient AI Models and Data Utilization

Linfeng Zhang got his Bachelor degree in Northeastern University and then got his Ph.D. degree in Tsinghua Univeristy. Currently, he leads Efficient and Precision Intelligent Computing (EPIC) lab in Shanghai Jiaotong Univeristy.

Shaobo Wang
Research Interest: Efficient Data-Centric AI
Contact: shaobowang1009@sjtu.edu.cn, gszfwsb.github.io

Shaobo Wang is a Ph.D. candidate in the EPIC Lab at SAI, Shanghai Jiao Tong University, starting in 2024. Building on a strong background in efficient AI, explainable AI, and deep learning theory, he focuses his research on data synthesis and data reduction. He is particularly interested in foundation models, striving to understand their intrinsic behavior while making them more data efficient, lightweight, and cost effective in both training and inference.

Yifeng Gao
Research Interest: Efficient LLM
Contact: yifenggao.cn@gmail.com

Yifeng Gao is a master student in EPIC Lab, Shanghai Jiaotong University. His research interests focus on developing capable, reliable and efficient AI with algorithm and computing co-design. Currently, he focus on efficient inference of the multi-step reasoning on large language models as well as their truthworthiness.

Zichen Wen
Research Interest: Efficient Multi-Modal LLM
Contact: Zichen.Wen@outlook.com

Zichen Wen is a Ph.D. student in the EPIC Lab at Shanghai Jiao Tong University, under the supervision of Prof. Linfeng Zhang. He holds a B.S. degree in Computer Science from the University of Electronic Science and Technology of China. During his undergraduate studies, he published multiple research papers in prestigious AI conferences, including AAAI, ACM MM，etc. His research interests lie in Efficient Multi-Modal Large Models and Trustworthy AI, focusing on advancing the efficiency, reliability, and ethical aspects of artificial intelligence systems.

Xuelin Li
Research Interest: Efficient LLM
Contact: lxl.curiosity@gmail.com

Xuelin Li will begin pursuing a Ph.D. degree at the EPIC Lab in 2025. He is expected to graduate with a Bachelor's degree from the University of Electronic Science and Technology of China (UESTC), where he achieved a perfect GPA: 4.0/4.0 in all courses within the School of Software. During his undergraduate studies, he received numerous awards, including National Scholarship. His research interests focus on developing efficient inference paradigms for trustworthy multimodal large language models..

Zexuan Yan
Research Interest: Efficient multimodal and AIGC models
Contact: yzx_ustc@mail.ustc.edu.cn

Zexuan Yan is currently a senior student majoring in Computer Science and Technology at the University of Science and Technology of China, and will join the EPIC Lab of Zhang Linfeng's research group at the School of Artificial Intelligence, Shanghai Jiao Tong University in the fall of 2025. His research interests include multimodal models, AIGC, and diffusion model acceleration.

Zou Chang
Research Interest: Efficient Image and Video
Contact: https://github.com/Shenyi-Z/ToCa

Chang Zou is currently an undergraduate student at Yingcai Honors College, University of Electronic Science and Technology of China (UESTC), expected to complete his bachelor's degree in 2026. Originally from Chengdu, Sichuan, he doesn’t eat spicy food despite his hometown’s reputation. His primary research focus is on the efficient acceleration of AIGC, particularly Diffusion Models, and he has a solid background in mathematics and physics. In 2024, he began his internship at the EPIC Lab, where, under the guidance of his advisor, Linfeng Zhang, he contributed to submissions for ICLR and CVPR.

Xuyang Liu
Research Interest: Token-wise Acceleration for MLLM
Contact: https://xuyang-liu16.github.io/

Xuyang Liu is currently pursuing his M.S. degree at the College of Electronics and Information Engineering, Sichuan University. He is also a research intern at Taobao & Tmall Group, where he focuses on efficient multi-modal large language models. In 2024, he joined the EPIC Lab as a research intern under the guidance of Prof. Linfeng Zhang, contributing to the development of a comprehensive collection of resources on token-level model compression. His research interests include Efficient AI, covering areas such as discrimination, adaptation, reconstruction, and generation.

Publication and Preprints
LEGION: Learning to Ground and Explain for Synthetic Image Detection
Hengrui Kang, Siwei Wen, Zichen Wen, Junyan Ye, Weijia Li, Peilin Feng, Baichuan Zhou, Bin Wang, Dahua Lin, Linfeng Zhang, Conghui He
arXiv preprint arXiv:2503.15264
LazyMAR: Accelerating Masked Autoregressive Models via Feature Caching
Feihong Yan, Qingyan Wei, Jiayi Tang, Jiajun Li, Yulin Wang, Xuming Hu, Huiqi Li, Linfeng Zhang
arXiv preprint arXiv:2503.12450
EEdit: Rethinking the Spatial and Temporal Redundancy for Efficient Image Editing
Zexuan Yan, Yue Ma, Chang Zou, Wenteng Chen, Qifeng Chen, Linfeng Zhang
arXiv preprint arXiv:2503.10270
OmniSAM: Omnidirectional Segment Anything Model for UDA in Panoramic Semantic Segmentation
Ding Zhong, Xu Zheng, Chenfei Liao, Yuanhuiyi Lyu, Jialei Chen, Shengyang Wu, Linfeng Zhang, Xuming Hu
arXiv preprint arXiv:2503.07098
From Reusing to Forecasting: Accelerating Diffusion Models with TaylorSeers
Jiacheng Liu, Chang Zou, Yuanhuiyi Lyu, Junjie Chen, Linfeng Zhang
arXiv preprint arXiv:2503.06923
ProReflow: Progressive Reflow with Decomposed Velocity
Lei Ke, Haohang Xu, Xuefei Ning, Yu Li, Jiajun Li, Haoling Li, Yuxuan Lin, Dongsheng Jiang, Yujiu Yang, Linfeng Zhang
arXiv preprint arXiv:2503.04824
Dataset Distillation with Neural Characteristic Function: A Minmax Perspective
Shaobo Wang, Yicun Yang, Zhiyuan Liu, Chenghao Sun, Xuming Hu, Conghui He, Linfeng Zhang
arXiv preprint arXiv:2502.20653
Token Pruning in Multimodal Large Language Models: Are We Solving the Right Problem?
Zichen Wen, Yifeng Gao, Weijia Li, Conghui He, Linfeng Zhang
arXiv preprint arXiv:2502.11501
Stop Looking for “Important Tokens” in Multimodal Language Models: Duplication Matters More
Zichen Wen, Yifeng Gao, Shaobo Wang, Junyuan Zhang, Qintong Zhang, Weijia Li, Conghui He, Linfeng Zhang
arXiv preprint arXiv:2502.11494
RealRAG: Retrieval-augmented Realistic Image Generation via Self-reflective Contrastive Learning
Yuanhuiyi Lyu, Xu Zheng, Lutao Jiang, Yibo Yan, Xin Zou, Huiyu Zhou, Linfeng Zhang, Xuming Hu
arXiv preprint arXiv:2502.00848
Global Compression Commander: Plug-and-Play Inference Acceleration for High-Resolution Large Vision-Language Models
Xuyang Liu, Ziming Wang, Yuhang Han, Yingyao Wang, Jiale Yuan, Jun Song, Bo Zheng, Linfeng Zhang, Siteng Huang, Honggang Chen
arXiv preprint arXiv:2501.05179
Token Pruning for Caching Better: 9× Acceleration on Stable Diffusion for Free
Evelyn Zhang, Bang Xiao, Jiayi Tang, Qianli Ma, Chang Zou, Xuefei Ning, Xuming Hu, Linfeng Zhang
arXiv preprint arXiv:2501.00375
Accelerating Diffusion Transformers with Dual Feature Caching
Chang Zou, Evelyn Zhang, Runlin Guo, Haohang Xu, Conghui He, Xuming Hu, Linfeng Zhang
arXiv preprint arXiv:2412.18911
Learning Robust Anymodal Segmentor with Unimodal and Cross-modal Distillation.
Xu Zheng, Haiwei Xue, Jialei Chen, Yibo Yan, Lutao Jiang, Yuanhuiyi Lyu, Kailun Yang,Linfeng Zhang, Xuming Hu
arXiv preprint arXiv:2411.17141
Multi-Stage Vision Token Dropping: Towards Efficient Multimodal Large Language Model.
Ting Liu, Liangtao Shi, Richang Hong, Yue Hu, Quanjun Yin,Linfeng Zhang
arXiv preprint arXiv:2411.10803
Gnothi Seauton: Empowering Faithful Self-Interpretability in Black-Box Models.
Shaobo Wang, Hongxuan Tang, Mingyang Wang, Hongrui Zhang, Xuyang Liu, Weiya Li, Xuming Hu, Linfeng Zhang
arXiv preprint arXiv:2410.21815
Reef: Representation encoding fingerprints for large language models.
Jie Zhang, Dongrui Liu, Chen Qian, Linfeng Zhang , Yong Liu, Yu Qiao, Jing Shao
arXiv preprint arXiv:2410.14273
Decouple-Then-Merge: Towards Better Training for Diffusion Models.
Qianli Ma, Xuefei Ning, Dongrui Liu, Li Niu, Linfeng Zhang
arXiv preprint arXiv:2410.06664
Accelerating Diffusion Transformers with Token-wise Feature Caching.
Chang Zou, Xuyang Liu, Ting Liu, Siteng Huang, Linfeng Zhang
arXiv preprint arXiv:2410.05317
Accelerating Diffusion Models with One-to-Many Knowledge Distillation.
Linfeng Zhang, Kaisheng Ma
arXiv preprint arXiv:2410.04191
DRUPI: Dataset Reduction Using Privileged Information.
Shaobo Wang, Yantai Yang, Shuaiyu Zhang, Chenghao Sun, Weiya Li, Xuming Hu, Linfeng Zhang
arXiv preprint arXiv:2410.01611
Ditfastattn: Attention compression for diffusion transformer models.
Zhihang Yuan, Hanling Zhang, Pu Lu, Xuefei Ning, Linfeng Zhang, Tianchen Zhao, Shengen Yan, Guohao Dai, Yu Wang
Neural Information Processing Systems (NeurIPS2024).
Not all samples should be utilized equally: Towards understanding and improving dataset distillation.
Shaobo Wang, Yantai Yang, Qilong Wang, Kaixin Li, Linfeng Zhang, Linfeng Zhang, Junchi Yan
arXiv preprint arXiv:2408.12483
Be Your Own Teacher: Improve the Performance of Convolutional Neural Networks via Self Distillation.
Linfeng Zhang, Jiebo Song, Anni Gao, Jingwei Chen, Chenglong Bao, and Kaisheng Ma
IEEE International Conference on Computer Vision (ICCV2019).
SCAN: A Scalable Neural Networks Framework Towards Compact and Efficient Models
Linfeng Zhang, , Zhanhong Tan, Jiebo Song, Jingwei Chen, Chenglong Bao, and Kaisheng Ma.
Neural Information Processing Systems (NeurIPS2019)
Auxiliary Training: Towards Accurate and Robust Models
Linfeng Zhang, Muzhou Yu, Tong Chen, Zuoqiang Shi, Chenglong Bao, and Kaisheng Ma
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR2020)
Task-Oriented Feature Distillation
Linfeng Zhang, Yukang Shi, Zuoqiang Shi, Kaisheng Ma, and Chenglong Bao
Neural Information Processing Systems (NeurIPS2020).
Self-Distillation: Towards Efficient and Compact Neural Networks
Linfeng Zhang, Chenglong Bao, and Kaisheng Ma
IEEE Transactions of Pattern Analysis and Machine Intelligence (IEEE TPAMI)
Improve Object Detection with Feature-based Knowledge Distillation: Towards Accurate and Efficient Detectors
Linfeng Zhang, Kaisheng Ma
The International Conference on Learning Representations (ICLR2021)
Structured Knowledge Distillation for Accurate and Efficient Object Detection
Linfeng Zhang, Kaisheng Ma
IEEE Transactions of Pattern Analysis and Machine Intelligence (IEEE TPAMI)
Wavelet Knowledge Distillation: Towards Efficient Image-to-Image Translation
Linfeng Zhang, Xin Chen, Xiaobing Tu, Pengfei Wan, Ning Xu, Kaisheng Ma
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR2022)
Contrastive Deep Supervision
Linfeng Zhang, Xin Chen, Junbo Zhang, Runpei Dong, Kaisheng Ma
European Conference on Computer Vision Oral Presentation
Pointdistiller: structured knowledge distillation towards efficient and compact 3d detection
Linfeng Zhang, Runpei Dong, Huang-Shuo Tai Kaisheng Ma
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR2023)
A Good Data Augmentation Policy Is Not All You Need: A Multi-Task Learning Perspective
Linfeng Zhang, Kaisheng Ma
IEEE Transactions on Circuits and Systems for Video Technology (IEEE TCSVT)
Fine-grained emotion classification of Chinese microblogs based on graph convolution networks
Yuni Lai, Linfeng Zhang(equal contribution), Donghong Han, Rui Zhou, Guoren Wang
World Wide Web Journal
Finding the Task-Optimal Low-Bit Sub-Distribution in Deep Neural Networks
Runpei Dong, Zhanhong Tan, Mengdi Wu, Linfeng Zhang, Kaisheng Ma
International Conference on Machine Learning (ICML2022)
Autoencoders as Cross-Modal Teachers: Can Pretrained 2D Image Transformers Help 3D Representation Learning?
Runpei Dong, Zekun Qi, Linfeng Zhang, Junbo Zhang, Jianjian Sun, Zheng Ge, Li Yi, Kaisheng Ma.
International Conference on Learning Representation (ICLR2023)
Non-Structured DNN Weight Pruning -- Is It Beneficial in Any Platform?
Xiaolong Ma, Sheng Lin, Shaokai Ye, Zhezhi He, Linfeng Zhang, Geng Yuan, Sia Huat, Tan, Zhengang Li, Deliang Fan, Xuehai Qian, Xue Lin, Kaisheng Ma, Yanzhi Wang
IEEE Transactions on Neural Networks and Learning Systems (IEEE TNNLS)
SMART: screen-based gesture recognition on commodity mobile devices
Zimo Liao, Zhicheng Luo, Qianyi Huang, Linfeng Zhang, Fan Wu, Qian Zhang, Yi Wang.
Annual International Conference on Mobile Computing and Networking (MobiCom21), Oral
*equal contribution
Wavelet J-Net: A Frequency Perceptive on Convolutional Neural Networks
Linfeng Zhang, Xiaoman Zhang, Chenglong Bao, Kaisheng Ma
International Joint Conference on Neural Networks (IJCNN2021)
CORSD: Class-Oriented Relational Self Distillation
Muzhou Yu, Sia Huat Tan, Kailu Wu, Runpei Dong, Linfeng Zhang, Kaisheng Ma
IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP2023)
Structured Knowledge Distillation Towards Efficient and Compact Multi-View 3D Detection
Linfeng Zhang, Yukang Shi, Hung-Shuo Tai, Zhipeng Zhang, Yuan He, Ke Wang, Kaisheng Ma
British Machine Visual Conference (BMVC2024, Oral)
Region-aware knowledge distillation for efficient image-to-image translation
Linfeng Zhang, Xin Chen, Runpei Dong, Kaisheng Ma
British Machine Visual Conference (BMVC2024)
Tiny Updater: Towards Efficient Neural Network-Driven Software Updating
Linfeng Zhang, Kaisheng Ma, IEEE International Conference on Computer Vision (ICCV2023)
IEEE International Conference on Computer Vision (ICCV2023)
Multi-Frequency Representation with Privilege Information for Video Super-Resolution
Fei Li, Linfeng Zhang, Zikun Liu, Juan Lei, Zhenbo Li, Zhenbo Li.
IEEE International Conference on Computer Vision (ICCV2023)
Ada3D: Exploiting the Spatial Redundancy with Adaptive Inference for Efficient 3D Object Detection
Tianchen Zhao, Xuefei Ning, Ke Hong, Zhongyuan Qiu, Lu Pu, Linfeng Zhang, Yali Zhao, Lipu Zhou, Guohao Dai, Huazhong ang, Yu Wang.
IEEE International Conference on Computer Vision (ICCV2023)
Gesture Recognition Using Visible Light on Mobile Devices
Zimo Liao, Zhicheng Luo, Qianyi Huang, Linfeng Zhang, Fan Wu, Qian Zhang, Guihai Chen.
IEEE/ACM Transactions on Networking (IEEE/ACM TON)