PhD student at KIT, part of the AGL Lab. Martin finished his Master’s at the University of Freiburg before joining André in Karlsruhe.
His research looks at continuous contextual variation and contextualization of multi-agent reinforcement learning. Recent work shows how continuously shifting environmental and task parameters affect training dynamics, sample efficiency, and an agent’s ability to generalize across different rates of contextual change.
Publications
2026
Dynamic Context Scheduling: Learning Beyond the Static Universe
We study dynamic context scheduling as a training instrument for contextual reinforcement learning. Rather than treating intra-episode context variation as a deployment reality, we treat it as a controlled shaping mechanism. Thereby, context evolves within each training episode according to a predetermined schedule, exposing the policy to a richer and more temporally structured region of the environment parameter space. We introduce DYNAMICCARLENV, a framework that wraps contextual environments with pluggable schedule families, such as sinusoidal offsets or cosine annealing. Across CartPole, BipedalWalker and VehicleRacing with CARL contextualization, we show that dynamic schedules match or outperform static context baselines in the out-of-distribution (OOD) regimes. Interestingly, for the more complex BipedalWalker and VehicleRacing environments we also achieve higher in-distribution (ID) evaluation performance. Preliminary findings indicate that automatic search for multi-stage curricula can successfully discover schedules that improve generalization, performing comparably to extensive grid search over single-stage schedulers.
@article{mraz-arxiv26a,title={Dynamic Context Scheduling: Learning Beyond the Static Universe},author={Mráz, Martin and Biedenkapp, André},year={2026},journal={arXiv:2608.20799},}
