Yisen Wang
研究员/助理教授
Researches machine learning and deep learning, including adversarial robustness, graph learning and weakly supervised learning.
Yisen Wang's lab at Peking works on trustworthy & safe ML and graph learning. 60 papers since 2023, 57 at top venues, mostly ICLR and NeurIPS. 14 PhD students. 6 alumni, most often to Huawei and Tencent. Applications close .
At a glance
- papers, 2023–2026
- 16/ year
- at T1 venues
- 95%
- PhD students
- 14
- MS students
- 0
- alumni listed
- 6
Papers by yearfrom the lab homepage, 2023–2026
- 14 papers2026 · T1 14 · collecting
- 15 papers2025 · T1 15
- 11 papers2024 · T1 10
- 20 papers2023 · T1 18
T1 venuesOtherVenues: ICLR, NeurIPS, ICML, EMNLP, CVPR
Research focuspaper titles that use each topic's words
- Graph learning24 papers
- supervised learning2 papers
- Trustworthy & safe ML1 paper
- weakly0 papers
- weakly supervised0 papers
Paper list
From the lab homepage.
202614 papers
- JMLR (T1)A Theoretical Framework for Masked Pretraining (MPT)
- ICLR (T1)Autoregressive Models Rival Diffusion Models at ANY-ORDER Generation
- EMNLP (T1)Decoding Large Language Diffusion Models with Foreseeing Movement
- ICLR (T1)Difficult Examples Hurt Unsupervised Contrastive Learning: A Theoretical Perspective
- ICML (T1)GRASP: Graph Reasoning via Agentic Solving and Probing of LLMs
- TPAMI (T1)Jailbreak and Guard Aligned Language Models with Only Few In-Context Demonstrations
- EMNLP (T1)On Assessing Adversarial Robustness of Latent Reasoning Large Language Models
- TPAMI (T1)On the Adversarial Transferability of Generalized "Skip Connections"
- ICLR (T1)On the Limits of Sparse Autoencoders: A Theoretical Framework and Reweighted Remedy
- ICLR (T1)SAE as a Crystal Ball: Interpretable Features Predict Cross-domain Transferability of LLMs without Training
- ICML (T1)Skill-Pro: Learning Reusable Skills from Experience via Non-Parametric PPO for LLM Agents
- ICML (T1)SSL4RL: Revisiting Self-supervised Learning as Intrinsic Reward for Visual-Language Reasoning
202515 papers
- JMLR (T1)An Augmentation Overlap Theory of Contrastive Learning
- ICML (T1)An Augmentation-Aware Theory for Self-Supervised Contrastive Learning
- ICLR (T1)Beyond Interpretability: The Gains of Feature Monosemanticity on Model Robustness
- NeurIPS (T1)Finding and Reactivating Post-Trained LLMs' Hidden Safety Mechanisms
- NeurIPS (T1)G1: Teaching LLMs to Reason on Graphs with Reinforcement Learning
- ICML (T1)Identifying and Understanding Cross-Class Features in Adversarial Training
- ICML (T1)Incorporating Arbitrary Matrix Group Equivariance into KANs
- NeurIPS (T1)Language Ranker: A Lightweight Ranking framework for LLM Decoding
- ICLR (T1)Leveraging Flatness to Improve Information-Theoretic Generalization Bounds for SGD
- ICML (T1)Long-Short Alignment for Effective Long-Context Modeling in LLMs
- ICLR (T1)Projection Head is Secretly an Information Bottleneck
- ICLR (T1)Rethinking Invariance in In-context Learning
202411 papers
- NeurIPS (T1)A Canonization Perspective on Invariant and Equivariant Learning
- NeurIPS (T1)A Theoretical Understanding of Self-Correction through In-context Alignment
- ICLR (T1)Do Generated Data Always Help Contrastive Learning?
- NeurIPS (T1)Fight Back Against Jailbreaking via Prompt Adversarial Tuning
- ICML (T1)Look Ahead or Look Around? A Theoretical Comparison Between Autoregressive and Masked Pretraining
- ICLR (T1)Non-negative Contrastive Learning
- ICLR (T1)On the Role of Discrete Tokenization in Visual Representation Learning
- ICML (T1)PID: Prompt-Independent Data Protection Against Latent Diffusion Models
- Neural NetworksSampling Complex Topology Structures for Spiking Neural Networks
- ICML (T1)TERD: A Unified Framework for Safeguarding Diffusion Models Against Backdoors
- NeurIPS (T1)Understanding the Role of Equivariance in Self-supervised Learning
202320 papers
- ICLR (T1)A Message Passing Perspective on Learning Dynamics of Contrastive Learning
- NeurIPS (T1)Adversarial Examples Are Not Real Features
- NeurIPS (T1)Architecture Matters: Uncovering Implicit Mechanisms in Graph Contrastive Learning
- ICLR (T1)ArCL: Enhancing Contrastive Learning with Augmentation-Robust Representations
- NeurIPS (T1)Balance, Imbalance, and Rebalance: Understanding Robust Overfitting from a Minimax Game Perspective
- CVPR (T1)CFA: Class-wise Calibrated Fair Adversarial Training
- ICLR (T1)ContraNorm: A Contrastive Learning Perspective on Oversmoothing and Beyond
- IEEE Transaction on Image ProcessingEquilibrium Image Denoising with Implicit Differentiation
- CVPR (T1)Generalist: Decoupling Natural and Robust Generalization
- NeurIPS (T1)GEQ: Gaussian Kernel Inspired Equilibrium Models
- NeurIPS (T1)Identifiable Contrastive Learning with Automatic Feature Importance Discovery
- NeurIPS (T1)Laplacian Canonization: A Minimalist Approach to Sign and Basis Invariant Spectral Embedding
Where alumni went6 listed, 33% to industry
Industry 2Academia 4
Most often to
- Huawei
- Tencent
Also to
- MIT
- The University of Melbourne
- University of Tokyo
Apply to PekingFall 2027
Computer Science PhD (international applicants)
- Round
- Main
- Opens
- —
- GRE
- See program page
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- USUCLA
Baharan Mirzasoleiman
- Trustworthy & safe ML
- USMichigan
Walter Dempsey
- Learning theory & optimization
- USMIT
Stephen Bates
- Trustworthy & safe ML
- USCMU
Sarah Cen
- Learning theory & optimization
- GBImperial
Alessandra Russo
- Reasoning & neuro-symbolic AI
- USGeorgia Tech
Ashwin Pananjady
- 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.
Pekingcloses