Linfeng Zhang
助理教授
Works on compression and acceleration of AI models and data, including self-distillation.
Linfeng Zhang's lab at SJTU works on efficient training & inference. 89 papers since 2023, 71 at top venues, mostly ICLR and CVPR. Applications close .
At a glance
- papers, 2023–2026
- 23/ year
- at T1 venues
- 80%
- PhD students
- —
- MS students
- —
- alumni listed
- —
Papers by yearfrom DBLP, 2023–2026
- 49 papers2026 · T1 41 · collecting
- 27 papers2025 · T1 22
- 2 papers2024 · T1 1
- 11 papers2023 · T1 7
T1 venuesOtherVenues: ICLR, CVPR, ICCV, AAAI, ACL
Research focuspaper titles that use each topic's words
- Efficient training & inference24 papers
- compression8 papers
- acceleration5 papers
- artificial intelligence0 papers
- artificial0 papers
Paper list
From DBLP.
202649 papers
- ICLR (T1)Accelerating Diffusion Large Language Models with SlowFast Sampling: The Three Golden Principles (opens in a new tab)
- ECCV (T1)Accelerating Diffusion Transformers with Gaussian Process Rectified Feature Cache (opens in a new tab)
- CVPR (T1)Accelerating Diffusion-based Video Editing via Heterogeneous Caching: Beyond Full Computing at Sampled Denoising Timestep (opens in a new tab)
- CVPR (T1)Accelerating Streaming Video Large Language Models via Hierarchical Token Compression (opens in a new tab)
- ACL (T1)AgentSlimming: Towards Efficient and Cost-Aware Multi-Agent Systems (opens in a new tab)
- ACL (T1)Are We Using the Right Benchmark: An Evaluation Framework for Visual Token Compression Methods (opens in a new tab)
- ECCV (T1)AViTS: Adaptive Spatiotemporal Token Selection for Efficient Dynamic-Resolution Generation (opens in a new tab)
- CVPR (T1)Beyond Fixed Formulas: Data-Driven Linear Predictor for Efficient Diffusion Models (opens in a new tab)
- ECCV (T1)Bridging Visual Representation and Reinforcement Learning from Verifiable Rewards in Large Vision-Language Models (opens in a new tab)
- Findings of ACLBypassing Neural Evaluations for Fast Audio Editing via Adaptive Trajectory Extrapolation (opens in a new tab)
- AAAI (T1)D²Pruner: Debiased Importance and Structural Diversity for MLLM Token Pruning (opens in a new tab)
- CVPR (T1)DisCa: Accelerating Video Diffusion Transformers with Distillation-Compatible Learnable Feature Caching (opens in a new tab)
202527 papers
- ICLR (T1)Accelerating Diffusion Transformers with Token-wise Feature Caching (opens in a new tab)
- ACM MMCompute Only 16 Tokens in One Timestep: Accelerating Diffusion Transformers with Cluster-Driven Feature Caching (opens in a new tab)
- ACL (T1)Data Whisperer: Efficient Data Selection for Task-Specific LLM Fine-Tuning via Few-Shot In-Context Learning (opens in a new tab)
- CVPR (T1)Dataset Distillation with Neural Characteristic Function: A Minmax Perspective (opens in a new tab)
- CVPR (T1)Decouple-Then-Merge: Finetune Diffusion Models as Multi-Task Learning (opens in a new tab)
- NeurIPS (T1)Don't Just Chase "Highlighted Tokens" in MLLMs: Revisiting Visual Holistic Context Retention (opens in a new tab)
- ICCV (T1)EEdit ⚡: Rethinking the Spatial and Temporal Redundancy for Efficient Image Editing (opens in a new tab)
- NeurIPS (T1)Efficient Multi-modal Large Language Models via Progressive Consistency Distillation (opens in a new tab)
- NeurIPS (T1)EfficientVLA: Training-Free Acceleration and Compression for Vision-Language-Action Models (opens in a new tab)
- ICCV (T1)From Reusing to Forecasting: Accelerating Diffusion Models With Taylorseers (opens in a new tab)
- ICLR (T1)Gnothi Seauton: Empowering Faithful Self-Interpretability in Black-Box Transformers (opens in a new tab)
- EMNLP (T1)GraphKV: Breaking the Static Selection Paradigm with Graph-Based KV Cache Eviction (opens in a new tab)
20242 papers
202311 papers
- IEEE Trans. Circuits Syst. Video Technol.A Good Data Augmentation Policy is not All You Need: A Multi-Task Learning Perspective (opens in a new tab)
- ICCV (T1)Ada3D : Exploiting the Spatial Redundancy with Adaptive Inference for Efficient 3D Object Detection (opens in a new tab)
- ICLR (T1)Autoencoders as Cross-Modal Teachers: Can Pretrained 2D Image Transformers Help 3D Representation Learning? (opens in a new tab)
- ICASSP (T1)CORSD: Class-Oriented Relational Self Distillation (opens in a new tab)
- ICCV (T1)Multi-Frequency Representation Enhancement with Privilege Information for Video Super-Resolution (opens in a new tab)
- CVPR (T1)PointDistiller: Structured Knowledge Distillation Towards Efficient and Compact 3D Detection (opens in a new tab)
- BMVCRegion-aware Knowledge Distillation for Efficient Image-to-Image Translation (opens in a new tab)
- IJCNNRevisiting Data Augmentation in Model Compression: An Empirical and Comprehensive Study (opens in a new tab)
- TPAMI (T1)Structured Knowledge Distillation for Accurate and Efficient Object Detection (opens in a new tab)
- BMVCStructured Knowledge Distillation Towards Efficient Multi-View 3D Object Detection (opens in a new tab)
- ICCV (T1)Tiny Updater: Towards Efficient Neural Network-Driven Software Updating (opens in a new tab)
Where alumni went
Alumni are not listed on the lab homepage.
Apply to SJTUFall 2027
Computer Science PhD (international applicants)
- Round
- Self-funded
- Opens
- —
- GRE
- See program page
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- CNZhejiang
Yingming Li
- Data management & big data
- CNSJTU
Weiran Huang
- Efficient training & inference
- USNYU
Juliana Freire
- Recommendation, search & web
- CAToronto
Scott Sanner
- Reinforcement learning
- USUT Austin
Radu Marculescu
- Trustworthy & safe ML
- DETUM
Xiao Xiang Zhu
- Physics, climate & scientific computing
Your notes
Save this lab to keep notes.
Data
Some numbers could not be verified:
- The member list could not be found on the lab site.
- The alumni list could not be found on the lab site.
- The lab site and the paper record do not agree.
- Collected
- Verified
- Papers
- DBLP
- Members
- Not collected
- Alumni
- Not collected
Tiers are venue classifications, not lab rankings.
SJTUcloses