Fabian Fumagalli

Research Focus

My research focuses on eXplainable AI (XAI) and I am passionate about making machine learning models more transparent and trustworthy. By leveraging mathematical concepts such as Shapley interactions, which address limitations of the Shapley value, and functional ANOVA, I aim to make model decisions more understandable to users. My goal is to narrow the gap between complex machine learning models and practical, interpretable solutions.

Beyond feature-based explanations, I am exploring novel applications of the Shapley value and interactions to enhance other areas in machine learning, such as large language model (LLM) prompt composition and hyperparameter optimization.

I actively contribute to the development of shapiq, which extends the popular shap package to support any-order feature interactions. shapiq decouples the computation of game-theoretic concepts from feature-based explanations, enabling the application of Shapley values and interactions across various machine learning tasks.

Another area of my research focuses on explanations in dynamic environments, particularly in cases with distribution shifts and rapidly changing models, such as evolving data streams. In this context, I co-organized the TempXAI: Explainable AI for Time Series and Data Streams Tutorial-Workshop at ECML PKDD 2024 and the DynXAI: Explainable Artificial Intelligence from Static to Dynamic Workshop at ECML PKDD 2023.

I was part of the collaborative research centre TRR 318 Constructing Explainability.

latest posts

Dec 06, 2024	New Paths in Explainable AI with Shapley Interactions
Dec 03, 2024	What Are Shapley Interactions, and Why Should You Care?
Sep 18, 2023	Best Paper Award for TRR Researchers at xAI 2023

Selected Publications

NeurIPS 2025

Explaining Similarity in Vision-Language Encoders with Weighted Banzhaf Interactions

Hubert Baniecki, Maximilian Muschalik, Fabian Fumagalli, Barbara Hammer, Eyke Hüllermeier, and Przemyslaw Biecek

In Proceedings of Advances in Neural Information Processing Systems (NeurIPS), 2025
AAAI 2026 (Oral)

HyperSHAP: Shapley Values and Interactions for Hyperparameter Importance

Marcel Wever, Maximilian Muschalik, Fabian Fumagalli, and Marius Lindauer

In Fortieth AAAI Conference on Artificial Intelligence, (AAAI), 2025
EMNLP 2025

Investigating the Impact of Conceptual Metaphors on LLM-based NLI through Shapley Interactions

Meghdut Sengupta, Maximilian Muschalik, Fabian Fumagalli, Barbara Hammer, Eyke Hüllermeier, Debanjan Ghosh, and Henning Wachsmuth

In Findings of the Association for Computational Linguistics: EMNLP 2025, 2025

NAACL 2025 (Oral)

Adaptive Prompting: Ad-hoc Prompt Composition for Social Bias Detection

Maximilian Spliethöver, Tim Knebler, Fabian Fumagalli, Maximilian Muschalik, Barbara Eva Hammer, Eyke Hüllermeier, and Henning Wachsmuth

In The 2025 Annual Conference of the Nations of the Americas Chapter of the ACL, 2025

Bib

@inproceedings{anonymous2025adaptive,
  title = {Adaptive Prompting: Ad-hoc Prompt Composition for Social Bias Detection},
  author = {Splieth\"{o}ver, Maximilian and Knebler, Tim and Fumagalli, Fabian and Muschalik, Maximilian and Hammer, Barbara Eva and H\"{u}llermeier, Eyke and Wachsmuth, Henning},
  booktitle = {The 2025 Annual Conference of the Nations of the Americas Chapter of the ACL},
  year = {2025},
  url = {https://aclanthology.org/2025.naacl-long.122/},
}

ICLR 2025

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, 2025

Bib

@inproceedings{fumagalli2025exact,
  title = {Exact Computation of Any-Order Shapley Interactions for Graph Neural Networks},
  author = {Muschalik, Maximilian and Fumagalli, Fabian and Frazzetto, Paolo and Strotherm, Janine and Hermes, Luca and Sperduti, Alessandro and H{\"u}llermeier, Eyke and Hammer, Barbara},
  booktitle = {The Thirteenth International Conference on Learning Representations},
  year = {2025},
  url = {https://openreview.net/forum?id=9tKC0YM8sX},
}

AISTATS 2025

Unifying Feature-Based Explanations with Functional ANOVA and Cooperative Game Theory

Fabian Fumagalli, Maximilian Muschalik, Eyke Hüllermeier, Barbara Hammer, and Julia Herbinger

In The 28th International Conference on Artificial Intelligence and Statistics, 2025

Bib Website

@inproceedings{fumagalli2024unifyingfeaturebasedexplanationsfunctional,
  title = {Unifying Feature-Based Explanations with Functional ANOVA and Cooperative Game Theory},
  author = {Fumagalli, Fabian and Muschalik, Maximilian and Hüllermeier, Eyke and Hammer, Barbara and Herbinger, Julia},
  booktitle = {The 28th International Conference on Artificial Intelligence and Statistics},
  year = {2025},
}

NeurIPS 2024

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 (NeurIPS) Datasets and Benchmarks Track, 2024

Bib Poster Website

@inproceedings{muschalik2024shapiq,
  title = {shapiq: Shapley Interactions for Machine Learning},
  author = {Muschalik, Maximilian and Baniecki, Hubert and Fumagalli, Fabian and Kolpaczki, Patrick and Hammer, Barbara and H\"{u}llermeier, Eyke},
  booktitle = {The Thirty-eight Conference on Neural Information Processing Systems (NeurIPS) Datasets and Benchmarks Track},
  year = {2024},
  url = {https://openreview.net/forum?id=knxGmi6SJi},
}

ICML 2024

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

Bib Poster

@inproceedings{Fumagalli.2024,
  title = {Kernel{SHAP}-{IQ}: Weighted Least Square Optimization for Shapley Interactions},
  author = {Fumagalli, Fabian and Muschalik, Maximilian and Kolpaczki, Patrick and H{\"u}llermeier, Eyke and Hammer, Barbara},
  year = {2024},
  booktitle = {Forty-first International Conference on Machine Learning (ICML)},
  url = {https://openreview.net/forum?id=d5jXW2H4gg},
}

AAAI 2024

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

DOI Bib Poster

@inproceedings{Muschalik.2024,
  title = {Beyond TreeSHAP: Efficient Computation of Any-Order Shapley Interactions for Tree Ensembles},
  author = {Muschalik, Maximilian and Fumagalli, Fabian and Hammer, Barbara and H{\"{u}}llermeier, Eyke},
  year = {2024},
  booktitle = {Thirty-Eighth {AAAI} Conference on Artificial Intelligence, (AAAI)},
  publisher = {{AAAI} Press},
  pages = {14388--14396},
  doi = {10.1609/AAAI.V38I12.29225},
  url = {https://doi.org/10.1609/aaai.v38i12.29225},
}

AISTATS 2024

SVARM-IQ: Efficient Approximation of Any-order Shapley Interactions through Stratification

Patrick Kolpaczki, Maximilian Muschalik, Fabian Fumagalli, Barbara Hammer, and Eyke Hüllermeier

In International Conference on Artificial Intelligence and Statistics (AISTATS), 2024

Bib

@inproceedings{Kolpaczki.2024,
  title = {{SVARM-IQ:} Efficient Approximation of Any-order Shapley Interactions through Stratification},
  author = {Kolpaczki, Patrick and Muschalik, Maximilian and Fumagalli, Fabian and Hammer, Barbara and H{\"{u}}llermeier, Eyke},
  year = {2024},
  booktitle = {International Conference on Artificial Intelligence and Statistics (AISTATS)},
  publisher = {{PMLR}},
  series = {Proceedings of Machine Learning Research},
  volume = {238},
  pages = {3520--3528},
  url = {https://proceedings.mlr.press/v238/kolpaczki24a.html},
}

IJCNN 2024

No learning rates needed: Introducing SALSA - Stable Armijo Line Search Adaptation

Philip Kenneweg, Tristan Kenneweg, Fabian Fumagalli, and Barbara Hammer

In International Joint Conference on Neural Networks, (IJCNN), 2024

DOI Bib

@inproceedings{DBLP:conf/ijcnn/KennewegKFH24,
  author = {Kenneweg, Philip and Kenneweg, Tristan and Fumagalli, Fabian and Hammer, Barbara},
  title = {No learning rates needed: Introducing {SALSA} - Stable Armijo Line
                    Search Adaptation},
  year = {2024},
  booktitle = {International Joint Conference on Neural Networks, {(IJCNN)}},
  pages = {1--8},
  publisher = {{IEEE}},
  url = {https://doi.org/10.1109/IJCNN60899.2024.10650124},
  doi = {10.1109/IJCNN60899.2024.10650124},
}

NeurIPS 2023

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

Bib Poster

@inproceedings{Fumagalli.2023b,
  title = {{SHAP-IQ:} Unified Approximation of any-order Shapley Interactions},
  author = {Fumagalli, Fabian and Muschalik, Maximilian and Kolpaczki, Patrick and H{\"{u}}llermeier, Eyke and Hammer, Barbara},
  year = {2023},
  booktitle = {Advances in Neural Information Processing Systems 36 (NeurIPS)},
  url = {https://openreview.net/forum?id=IEMLNF4gK4},
}

MLJ

Incremental Permutation Feature Importance (iPFI): Towards Online Explanations on Data Streams

Fabian Fumagalli^*, Maximilian Muschalik^*, Eyke Hüllermeier, and Barbara Hammer

Machine Learning, 2023

DOI Bib Poster Slides

@article{Fumagalli.2023a,
  title = {Incremental Permutation Feature Importance (iPFI): Towards Online Explanations on Data Streams},
  author = {Fumagalli, Fabian and Muschalik, Maximilian and H{\"{u}}llermeier, Eyke and Hammer, Barbara},
  year = {2023},
  journal = {Machine Learning},
  volume = {112},
  number = {12},
  pages = {4863--4903},
  doi = {10.1007/S10994-023-06385-Y},
  url = {https://doi.org/10.1007/s10994-023-06385-y},
}

xAI 2023

iPDP: On Partial Dependence Plots in Dynamic Modeling Scenarios

Maximilian Muschalik^*, Fabian Fumagalli^*, Rohit Jagtani, Barbara Hammer, and Eyke Hüllermeier

In World Conference of Explainable Artificial Intelligence (xAI), 2023

DOI Bib Slides

@inproceedings{Muschalik.2023,
  title = {iPDP: On Partial Dependence Plots in Dynamic Modeling Scenarios},
  author = {Muschalik, Maximilian and Fumagalli, Fabian and Jagtani, Rohit and Hammer, Barbara and H{\"{u}}llermeier, Eyke},
  year = {2023},
  booktitle = {World Conference of Explainable Artificial Intelligence (xAI)},
  publisher = {Springer},
  series = {Communications in Computer and Information Science},
  volume = {1901},
  pages = {177--194},
  doi = {10.1007/978-3-031-44064-9\_11},
  url = {https://doi.org/10.1007/978-3-031-44064-9\_11},
}

ECML PKDD 2023

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

DOI Bib Poster Slides

@inproceedings{Muschalik.2023a,
  title = {iSAGE: An Incremental Version of {SAGE} for Online Explanation on Data Streams},
  author = {Muschalik, Maximilian and Fumagalli, Fabian and Hammer, Barbara and H{\"{u}}llermeier, Eyke},
  booktitle = {Machine Learning and Knowledge Discovery in Databases: Research Track - European Conference, ({ECML} {PKDD})},
  publisher = {Springer},
  series = {Lecture Notes in Computer Science},
  volume = {14171},
  year = {2023},
  pages = {428--445},
  doi = {10.1007/978-3-031-43418-1\_26},
  url = {https://doi.org/10.1007/978-3-031-43418-1\_26},
}

KI Journal

Agnostic Explanation of Model Change based on Feature Importance

Maximilian Muschalik^*, Fabian Fumagalli^*, Barbara Hammer, and Eyke Hüllermeier

Künstliche Intelligenz, 2022

DOI Bib Website

@article{Muschalik.2022,
  title = {Agnostic Explanation of Model Change based on Feature Importance},
  author = {Muschalik, Maximilian and Fumagalli, Fabian and Hammer, Barbara and H{\"{u}}llermeier, Eyke},
  year = {2022},
  journal = {K{\"{u}}nstliche Intelligenz},
  volume = {36},
  number = {3},
  pages = {211--224},
  doi = {10.1007/s13218-022-00766-6},
}