Stephan Günnemann
Professor
Makes machine learning reliable and robust, with a focus on graph neural networks, adversarial robustness, uncertainty estimation and machine learning for molecules.
Stephan Günnemann's lab at TUM works on trustworthy & safe ML and probabilistic & causal inference. 81 papers since 2023, 34 at top venues, mostly ICLR and ICML. 16 alumni.
At a glance
- papers, 2023–2026
- 21/ year
- at T1 venues
- 42%
- PhD students
- 0
- MS students
- 0
- alumni listed
- 16
Papers by yearfrom the lab homepage, 2023–2026
- 16 papers2026 · T1 13 · collecting
- 23 papers2025 · T1 20
- 18 papers2024 · T1 1
- 24 papers2023 · T1 0
T1 venuesOtherVenues: ICLR, ICML, NeurIPS, TMLR, ICCV
Research focuspaper titles that use each topic's words
- Trustworthy & safe ML7 papers
- Probabilistic & causal inference7 papers
- adversarial robustness6 papers
- molecules4 papers
- uncertainty estimation2 papers
Paper list
From the lab homepage.
202616 papers
- ICML (T1)A Coin Flip for Safety: LLM Judges Fail to Reliably Measure Adversarial Robustness
- Transactions of Machine Learning Research (TMLR), J2C-Certification - Presentation @ ICLRAdversarial Robustness of Graph Transformers
- TMLR (T1)Amplified Patch-Level Differential Privacy for Free via Random Cropping
- ICML (T1)Certifying Graph Neural Networks Against Label and Structure Poisoning
- ICML (T1)Derivative Informed Learning of Exchange-Correlation Functionals
- ICLR (T1)Discrete Bayesian Sample Inference for Graph Generation
- ICLR (T1)Edit-Based Flow Matching for Temporal Point Processes
- ECCV (T1)EgoDyn-Bench: Evaluating Ego-Motion Understanding in Vision-Centric Foundation Models for Autonomous Driving
- International Conference on Learning Representations, ICLR @ Trustworthy AI WorkshopExact Certification of Neural Networks and Partition Aggregation Ensembles against Label Poisoning
- ICLR (T1)Model Collapse Is Not a Bug but a Feature in Machine Unlearning for LLMs
- TMLR (T1)Predictive Feature Caching for Training-free Acceleration of Molecular Geometry Generation
- ICLR (T1)Sampling-aware Adversarial Attacks Against Large Language Models
202523 papers
- ICLR (T1)A Probabilistic Perspective on Unlearning and Alignment for Large Language Models
- Transactions of Machine Learning Research, TMLRA Unified Approach Towards Active Learning and Out-of-Distribution Detection
- ICLR (T1)Efficient Time Series Processing for Transformers and State-Space Models through Token Merging
- Knowledge and Information Systems, KAIS .Evaluating the transferability of adversarial robustness to target domains
- ICLR (T1)Exact Certification of (Graph) Neural Networks Against Label Poisoning
- ICLR (T1)Flow Matching with Gaussian Process Priors for Probabilistic Time Series Forecasting
- ICCV (T1)GeoDiffusion: A Training-Free Framework for Accurate 3D Geometric Conditioning in Image Generation
- ICLR (T1)Graph Neural Networks for Edge Signals: Orientation Equivariance and Invariance
- CVPR (T1)Joint Out-of-Distribution Filtering and Data Discovery Active Learning
- NeurIPS (T1)Joint Relational Database Generation via Graph-Conditional Diffusion Models
- ICLR (T1)Learning Equivariant Non-local Electron Density Functionals
- ICLR (T1)Lift Your Molecules: Molecular Graph Generation in Latent Euclidean Space
202418 papers
- TMLR (T1)Assessing Robustness via Score-Based Adversarial Image Generation
- Next Generation of AI Safety Workshop @ ICML .Attacking Large Language Models With Projected Gradient Descent
- IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)Deep Sensor Fusion with Constraint Safety Bounds for High Precision Localization
- Conference on Neural Information Processing Systems (NeurIPS)Efficient adversarial training in llms with continuous attacks
- Conference on Neural Information Processing Systems (NeurIPS)Energy-based Epistemic Uncertainty for Graph Neural Networks
- Conference on Neural Information Processing Systems (NeurIPS)Expected Probabilistic Hierarchies
- International Conference on Machine Learning (ICML), . Oral presentationExpressivity and Generalization: Fragment-Biases for Molecular GNNs
- International Conference on Learning Representations (ICLR)From Zero to Turbulence: Generative Modeling for 3D Flow Simulation
- IEEE Robotics and Automation Letters (RAL)Generalized Synchronized Active Learning for Multi-Agent-Based Data Selection on Mobile Robotic Systems
- Conference on Neural Information Processing Systems (NeurIPS)Neural Pfaffians: Solving Many Many-Electron Schrödinger Equations
- AI4DifferentialEquations in Science @ ICLROn Representing Electronic Wave Functions with Sign Equivariant Neural Networks
- Transactions of Machine Learning Research (TMLR), also Best Paper Award, AdvML-Frontiers @ NeurIPSProvable Robustness of (Graph) Neural Networks Against Data Poisoning and Backdoor Attacks
202324 papers
- Conference on Neural Information Processing Systems (NeurIPS)Add and Thin: Diffusion for Temporal Point Processes
- Conference on Neural Information Processing Systems (NeurIPS)Adversarial Training for Graph Neural Networks: Pitfalls, Solutions, and New Directions
- Workshop on Robotics and AI: The Future of Industrial Assembly Tasks, Robotics: Science and Systems (RSS)Density-based Feasibility Learning with Normalizing Flows for Introspective Robotic Assembly
- (best paper award) In AISafety & SafeRL Joint Workshop at IJCAIDiffusion Denoised Smoothing for Certified and Adversarial Robust Out-Of-Distribution Detection
- Quantum Computing and EngineeringEfficient MILP Decomposition in Quantum Computing for ReLU Network Robustness
- International Conference on Machine Learning (ICML)Ewald-based Long-Range Message Passing for Molecular Graphs
- Workshop on New Frontiers in Adversarial Machine Learning, ICMLExpressivity of Graph Neural Networks Through the Lens of Adversarial Robustness
- International Conference on Machine Learning (ICML)Generalizing Neural Wave Functions
- June edition of the IEEE Data Engineering BulletinGraph Data Augmentation for Graph Machine Learning: A Survey
- Conference on Neural Information Processing Systems (NeurIPS)Hierarchical Randomized Smoothing
- International Conference on Learning Representations (ICLR)Localized Randomized Smoothing for Collective Robustness Certification
- International Conference on Machine Learning (ICML)Modeling Temporal Data as Continuous Functions with Stochastic Process Diffusion
Where alumni went16 listed
Apply to TUMRolling
Not announced yet
Computer Science Doctorate (CIT)
- Round
- Rolling (position-based)
- Opens
- —
- GRE
- See program page
- English
- Very good written and spoken English; no minimum test score
- Fee
- See program page
Confirmed on the program page.
Similar labsnearest by research summary
- SGNUS
Anji Liu
- Probabilistic & causal inference
- CNPeking
Haoxuan Li
- Probabilistic & causal inference
- CNZhejiang
Xiaofei He
- Recommendation, search & web
- USUCSD
Yoav Freund
- Learning theory & optimization
- USUIUC
Jiaxuan You
- Graph learning
- CAToronto
Chris J. Maddison
- Probabilistic & causal inference
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.