van der Schaar Lab at ICML workshop 2026

We are delighted to share a variety of papers from the van der Schaar Lab have been accepted across multiple workshops at the International Conference on Machine Learning (ICML) 2026, taking place in Seoul, South Korea.

In addition to our papers accepted to the ICML 2026 main conference track, these workshop acceptances further highlight the breadth and depth of the lab’s research contributions this year.

Our accepted workshop papers are:

1. Scaling with Recursion in Masked Discrete Diffusion Models
   Alba Carballo Castro, Julianna Piskorz, Paulius Rauba, Mihaela van der Schaar
   Accepted at Foundations of Deep Generative Models, High Dimensional Learning Dynamics, and Structured Probabilistic Inference & Generative Modeling
2. A Decision Theoretic Formalisation of Steganography with Applications to LLM Monitoring
   Usman Anwar, Julianna Piskorz, David D. Baek, David D. Africa, Jim Weatherall, Max Tegmark, Christian Schroeder de Witt, Mihaela van der Schaar, David Krueger
   Accepted at Technical AI Governance Research
3. Knowledge-Informed Kernel State Reconstruction from Heterogeneous Partial Observations
   Luca Muscarnera, Silas Ruhrberg Estévez, Samuel Holt, Evgeny Saveliev, Mihaela van der Schaar
   Accepted at Structured Data for Health
4. Learning Reasoning Rewards from Expert Demonstrations with Inverse Reinforcement Learning
   Claudio Fanconi, Nicolas Astorga, Mihaela van der Schaar
   Accepted at Decision-Making from Offline Datasets to Online Adaption
5. Resource-Adaptive Foundation Model Reasoning via Semantic Coverage
   Max Ruiz Luyten, Thomas Pouplin, Mihaela van der Schaar
   Accepted at AdaptFM
6. Structural Support Certificates for Graph-Query Inference in Mechanism Posteriors
   Max Ruiz Luyten, Mihaela van der Schaar
   Accepted at Structured Probabilistic Inference & Generative Modeling
7. Communication Boundary Failures in Multi-Agent Language Models
   Max Ruiz Luyten, Mihaela van der Schaar
   Accepted at Failure Modes in Agentic AI
8. When More Rollouts Add No Support: Semantic Coverage Collapse in Agentic Test-Time Search
   Max Ruiz Luyten, Thomas Pouplin, Mihaela van der Schaar
   Accepted at Failure Modes in Agentic AI
9. Reasoning with Neologisms: Can Soft Tokens Learn Composable Reasoning Skills Without Forgetting?
   Antonin Berthon, Mihaela van der Schaar
   Accepted at Compositional Learning: Safety, Interpretability, and Agents