From Code to Action: Hierarchical Learning of Diffusion-VLM Policies
Markus Peschl, Pietro Mazzaglia, Daniel Dijkman,
IROS 2026 (International Conference on Intelligent Robots and Systems)
arXiv

We use VLMs and diffusion policies with API subroutines as structured supervision to improve generalization and long-horizon robotic control.

Hybrid Training for Vision-Language-Action Models
Pietro Mazzaglia, Cansu Sancaktar, Markus Peschl, Daniel Dijkman,
ICLR 2026 (International Conference on Learning Representations)
OpenReview / arXiv

Hybrid Training (HyT) lets Vision-Language-Action models learn from chain-of-thought reasoning to boost performance while allowing faster, thought-free inference when needed.

Focusing on What Matters: Object-Agent-centric Tokenization for Vision Language Action models
Rokas Bendikas, Daniel Dijkman, Markus Peschl, Sanjay Haresh, Pietro Mazzaglia,
CoRL 2025 (Conference on Robot Learning)
OpenReview

Introducing Oat-VLA: Cut visual tokens for VLAs and train twice as fast as OpenVLA while achieving better real-world performance.

Differentiable and Learnable Wireless Simulation with Geometric Transformers
Thomas Hehn, Markus Peschl, Tribhuvanesh Orekondy, Arash Behboodi, Johann Brehmer,
ICLR 2025 (International Conference on Learning Representations)
GitHub / arXiv

A Geometric Algebra Transformer "WiGATr" for learning an E(3) equivariant simulator of wireless signal transmission. We test WiGATr as a predictive model as well as a diffusion model of signal and 3D geometry.

NeuroSteiner: A Graph Transformer for Wirelength Estimation
Sahil Manchanda, Dana Kianfar, Markus Peschl, Romain Lepert, Michaël Defferrard,
4th AAAI Workshop on AI to Accelerate Science and Engineering
arXiv

Using graph transformers to predict Steiner points in a hybrid fashion to tackle the Steiner Tree problem in physical design (chip design).

Robust scheduling with GFlowNets
David W Zhang, Corrado Rainone, Markus Peschl, Roberto Bondesan,
ICLR 2023 (International Conference on Learning Representations)
OpenReview / arXiv

We use GFlowNets to tackle the NP-hard problem of computation graph scheduling by combining a temparature conditioned policy with top-k sampling.

Learning Perturbations for Soft-Output Linear MIMO Demappers
Daniel E. Worrall, Markus Peschl, Arash Behboodi, Roberto Bondesan,
GLOBECOM 2022 (IEEE Global Communications Conference)
IEEE / arXiv

We combine lattice reduction, Bayesian optimization and a stochastic sampler to arrive at a highly efficient, yet accurate linear MIMO demapper.

MORAL: Aligning AI with Human Norms through Multi-Objective Reinforced Active Learning
Markus Peschl, Arkady Zgonnikov, Frans A Oliehoek, Luciano C Siebert,
AAMAS 2022 (International Conference on Autonomous Agents and Multiagent Systems)
arXiv

AI alignment is challenging due to a multitude of different and possibly conflicting values present in human feedback. We propose to overcome this challenge with a multi-objective agent that actively learns user preferences.