Michael Mühlebach

Research Interests

• Foundations of machine learning

• Large-language models (inference-time compute, data selection, reasoning/agentic systems)

• Reinforcement learning and online decision-making

• Mathematical optimization for machine learning

• Control theory

Recent publications/preprints

• C. Vernade, O. Eberhard, M. White, F. Dörfler, C. Szepesvári, M. Krstic, M. Muehlebach, "Foundations of Reinforcement Learning and Control: Connections and New Perspectives," Tutorials in Operations Research, 2026, https://arxiv.org/abs/2608.02433

• H. Ma, M. Bal, L. Zhang, B. Li, N. He, M. Zeilinger, M. Muehlebach, "SALAAD: Sparse And Low-Rank Adaptation via ADMM for Large Language Model Inference," 2026, https://arxiv.org/abs/2602.00942

• M. Muehlebach, M. I. Jordan, "Accelerated First-Order Optimization under Nonlinear Constraints," Mathematical Programming, 2025,  https://arxiv.org/abs/2302.00316

• L. Zhang, B. Li, K. K. Thekumparampil, S. Oh, M. Muehlebach, N. He, "Zeroth-Order Minimization finds Flat Minima," Advances in Neural Information Processing Systems, 2025, https://arxiv.org/abs/2506.05454v2

• M. Muehlebach, Z. He, M. I. Jordan, "The Sample Complexity of Online Reinforcement Learning: A Multi-Model Perspective," International Conference on Learning Representations, 2026, https://arxiv.org/abs/2501.15910