Maximilian Muschalik

Artificial Intelligence and Machine Learning at LMU Munich.

prof_pic.jpg

Akademiestr. 7

80799 Munich

Hi, I am Maximilian Muschalik, and I am a postdoctoral researcher at Prof. Eyke Hüllermeier’s AIML chair, LMU Munich, and the Munich Center for Machine Learning (MCML). I received my PhD (summa cum laude) from LMU Munich in 2026. I also like Shapley interactions quite a lot. So I work on them and develop shapiq.

At NeurIPS 2026, we present probly: Uncertainty-Aware Machine Learning as an oral. It is my first paper as last author, and probly is an open-source Python package that brings many uncertainty quantification methods into one toolbox.

shapiq logoshapiq logo probly logoprobly logo

Research Focus

My research centers on the topic of explainable artificial intelligence (XAI), with a focus on black-box machine learning models.

Currently, I am mainly working on the topic of Shapley-based explanations and extensions of Shapley values to higher-order interactions. Therein, we are mostly concerned with developing approximation methods to compute Shapley interactions efficiently, as usually the computational complexity of Shapley interactions is infeasible for most real-world applications. We have bundled our research and methods into the Python package shapiq, which unifies state-of-the-art algorithms to efficiently compute Shapley values and any-order Shapley interactions in an application-agnostic framework. More recently, we have pushed Shapley-based explanations into new modalities, including graph neural networks, vision-language encoders, and hyperparameter optimization. Lately, I have also been working on uncertainty quantification, which we bundle in the Python package probly. If you are interested in Shapley interactions, I would be happy to hear from you. You can also check out our blog post on Shapley interactions.

Another interesting area of XAI research is the development of methods that can explain the predictions of models in dynamic learning environments. Specifically, we investigate the challenges of creating accurate and timely explanations for models that must constantly adapt to changes in data streams and learning tasks. In such dynamic settings, traditional XAI methods may be computationally expensive or unable to provide faithful explanations in a timely manner.

My research is part of the TRR 318 Constructing Explainability.

News

Nov 13, 2026 Our paper probly: Uncertainty-Aware Machine Learning, my first paper as last author, will be presented as an oral at NeurIPS 2026 in Sydney, together with Proxy-Based Approximation of Shapley and Banzhaf Interactions. :tada:
Jul 10, 2026 Our papers ShaplEIG: Bayesian Experimental Design for Shapley Value Estimation and Exactly Computing do-Shapley Values have been presented at ICML 2026 in Seoul.
Jun 30, 2026 I completed my PhD at LMU Munich summa cum laude. My thesis Feature-Based Explainable AI: From Attribution to Interaction, From Static to Dynamic is available here. :tada:
Apr 25, 2026 Our paper Efficient Credal Prediction through Decalibration has been presented at ICLR 2026.
Jan 20, 2026 Our paper HyperSHAP: Shapley Values and Interactions for Explaining Hyperparameter Optimization has been presented at AAAI 2026 in Singapore. :tada:

Latest Posts

Selected Publications

  1. NeurIPS
    Oral
    probly: Uncertainty-Aware Machine Learning
    Paul Hofman, Clemens Damke, Timo Löhr, Santo M. A. R. Thies, Alireza Javanmardi, Valentin Margraf, Jakub Paplhám, Yusuf Sale, Eyke Hüllermeier, and Maximilian Muschalik
    In Advances in Neural Information Processing Systems 39 (NeurIPS 2026), Evaluations and Datasets Track, 2026
  2. NeurIPS
    Proxy-Based Approximation of Shapley and Banzhaf Interactions
    Santo M. A. R. Thies, Hubert Baniecki, R. Teal Witter, Eyke Hüllermeier, Maximilian Muschalik†, and Fabian Fumagalli†
    In Advances in Neural Information Processing Systems 39 (NeurIPS 2026), 2026
    † Equal senior authorship
  3. NeurIPS
    Explaining Similarity in Vision-Language Encoders with Weighted Banzhaf Interactions
    Hubert Baniecki, Maximilian Muschalik, Fabian Fumagalli, Barbara Hammer, Eyke Hüllermeier, and Przemyslaw Biecek
    In Advances in Neural Information Processing Systems 38 (NeurIPS 2025), 2025
  4. ICLR
    Exact Computation of Any-Order Shapley Interactions for Graph Neural Networks
    Maximilian Muschalik, Fabian Fumagalli, Paolo Frazzetto, Janine Strotherm, Luca Hermes, Alessandro Sperduti, Eyke Hüllermeier, and Barbara Hammer
    In The Thirteenth International Conference on Learning Representations (ICLR 2025), 2025
  5. NeurIPS
    shapiq: Shapley Interactions for Machine Learning
    Maximilian Muschalik, Hubert Baniecki, Fabian Fumagalli, Patrick Kolpaczki, Barbara Hammer, and Eyke Hüllermeier
    In The Thirty-eight Conference on Neural Information Processing Systems Datasets and Benchmarks Track (NeurIPS 2024), 2024
  6. AAAI
    Beyond TreeSHAP: Efficient Computation of Any-Order Shapley Interactions for Tree Ensembles
    Maximilian Muschalik, Fabian Fumagalli, Barbara Hammer, and Eyke Hüllermeier
    In Thirty-Eighth AAAI Conference on Artificial Intelligence, (AAAI 2024), 2024
  7. ICML
    KernelSHAP-IQ: Weighted Least Square Optimization for Shapley Interactions
    Fabian Fumagalli, Maximilian Muschalik, Patrick Kolpaczki, Eyke Hüllermeier, and Barbara Hammer
    In Forty-first International Conference on Machine Learning (ICML 2024), 2024
  8. ECMLPKDD
    iSAGE: An Incremental Version of SAGE for Online Explanation on Data Streams
    Maximilian Muschalik, Fabian Fumagalli, Barbara Hammer, and Eyke Hüllermeier
    In Machine Learning and Knowledge Discovery in Databases: Research Track - European Conference, (ECML PKDD 2023), 2023
  9. NeurIPS
    SHAP-IQ: Unified Approximation of any-order Shapley Interactions
    Fabian Fumagalli, Maximilian Muschalik, Patrick Kolpaczki, Eyke Hüllermeier, and Barbara Hammer
    In Advances in Neural Information Processing Systems 36 (NeurIPS 2023), 2023