Quanshi Zhang
Tenured Associate Professor
Works on explainable AI, building an interaction-based theory of the representations of deep neural networks.
Quanshi Zhang's lab at SJTU works on learning theory & optimization, trustworthy & safe ML and human-AI collaboration & trust. 16 papers since 2023, 14 at top venues, mostly ICML and AAAI. 7 PhD students and 1 master's student. 20 alumni, most often to Ant Group and Huawei. Applications close .
At a glance
- papers, 2023–2026
- 4/ year
- at T1 venues
- 88%
- PhD students
- 7
- MS students
- 1
- alumni listed
- 20
Papers by yearfrom the lab homepage, 2023–2026
- 0 papers2026 · T1 0 · collecting
- 0 papers2025 · T1 0
- 7 papers2024 · T1 7
- 9 papers2023 · T1 7
T1 venuesOtherVenues: ICML, AAAI, ICLR, TPAMI, CVPR
Research focuspaper titles that use each topic's words
- Learning theory & optimization1 paper
- deep neural1 paper
- Trustworthy & safe ML0 papers
- Human-AI collaboration & trust0 papers
- explainable0 papers
Paper list
From the lab homepage.
20247 papers
- AAAI (T1)Batch Normalization Is Blind to the First and Second Derivatives of the Loss (opens in a new tab)
- AAAI (T1)Clarifying the Behavior and the Difficulty of Adversarial Training (opens in a new tab)
- ICLR (T1)Defining and Extracting generalizable interaction primitives from DNNs (opens in a new tab)
- AAAI (T1)Explaining Generalization Power of a DNN using Interactive Concepts
- TPAMI (T1)Interpretable Rotation-Equivariant Quaternion Neural Networks for 3D Point Cloud Processing
- TPAMI (T1)Unifying Fourteen Post-Hoc Attribution Methods With Taylor Interactions (opens in a new tab)
- ICLR (T1)Where We Have Arrived in Proving the Emergence of Sparse Interaction Primitives in AI Models (opens in a new tab)
20239 papers
- ICML (T1)Bayesian Neural Networks Avoid Encoding Perturbation-sensitive and Complex Concepts (opens in a new tab)
- ICLR (T1)Can We Faithfully Represent Masked States to Compute Shapley Values on a DNN? (opens in a new tab)
- ICML (T1)Defects of Convolutional Decoder Networks in Frequency Representation (opens in a new tab)
- CVPR (T1)Defining and Quantifying the Emergence of Sparse Concepts in DNNs (opens in a new tab)
- ICML (T1)Does a Neural Network Really Encode Symbolic Concept? (opens in a new tab)
- ICML (T1)HarsanyiNet: Computing Accurate Shapley Values in a Single Forward Propagation (opens in a new tab)
- Machine Intelligence ResearchInterpretability of Neural Networks Based on Game-Theoretic Interactions
- IEEE Transactions on Mobile ComputingLearning to Prevent Input Leakages in the Mobile Cloud Inference
- NeurIPS (T1)Towards the Difficulty for a Deep Neural Network to Learn Concepts of Different Complexities
Where alumni went20 listed, 11% to industry
Industry 2Academia 16
Most often to
- Ant Group
- Huawei
Also to
- UC Berkeley
- University of Toronto
- Carnegie Mellon University
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
- USStanford
Carlos Guestrin
- Trustworthy & safe ML
- CNSJTU
Nanyang Ye
- Learning theory & optimization
- CHEPFL
Florent Krzakala
- Learning theory & optimization
- SGNTU
Han Yu
- Trustworthy & safe ML
- SGNTU
Atsushi Nitanda
- Learning theory & optimization
- USGeorgia Tech
Jacob Abernethy
- Learning theory & optimization
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- The lab homepage
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
SJTUcloses