Martino Ciaperoni

I am a postdoctoral researcher at Scuola Normale Superiore in Pisa, in the Data Science and Human-centered AI group and the KDD Lab led by Prof. Fosca Giannotti. My current research focuses on explainable and trustworthy AI, the evaluation and explanation of large language model behavior, including behavioral shifts, and applications of AI to scientific discovery. I am a work package and task leader in the ERC Proof of Concept project ILLUME-4-Science.

Previously, I was a postdoctoral researcher at Aalto University with Prof. Heikki Mannila, and I received my D.Sc. (Tech.) in Computer Science from Aalto University in 2024, advised by Prof. Aristides Gionis. Before that, I studied statistics and data science at the University of Bologna, the University of Glasgow, and Sapienza University of Rome, and carried out research at ISI Foundation in Turin with Dr. Francesco Bonchi.

My work has appeared at venues including NeurIPS, KDD, SIGMOD, IJCAI, ICDE, ICDM and ECML PKDD, and in journals such as Data Mining and Knowledge Discovery and ACM Transactions on Knowledge Discovery from Data.

Research

Explainable & trustworthy AI

Making AI systems transparent, understandable, and reliable.

Behavioral shifts in LLMs

Understanding and explaining how the behavior of large language models changes.

AI for science

AI methods that support scientific discovery.

Interpretable machine learning

Machine learning models that are interpretable by design.

Algorithms for data

Efficient algorithms for data mining, data management, and machine learning.

News

Sep 2026	Position paper accepted at NeurIPS 2026: Comparing Explanations is Not Enough, Explain the Change: New Standards are Needed to Explain Behavioral Shifts in Large Language Models. arXiv
Sep 2026	At ECML PKDD 2026 in Naples: presented Virgil: Navigating Explainability for Transformer-based Language Models (oral and demo), chaired the session Large Language Models and AI in Biology and Medicine, and received an Outstanding Emergency Reviewer recognition (Applied Data Science Track).
Aug 2026	Paper at IJCAI 2026 (Bremen): Efficient Optimization of Fixed-Length Paths, with N. Tziavelis and P. Karras.
Aug 2026	Presented Explainable AI for Cancer Drug Response Prediction: Beyond Univariate Feature Attributions at KDD 2026 (Jeju Island).
May 2026	The ERC Proof of Concept project ILLUME-4-Science has started. I lead the work package on demonstrating scientific impact. Read more (in Italian)
Mar 2026	Paper at EDBT 2026 (Tampere): Khatri-Rao Clustering for Data Summarization, with C. Leiber, A. Gionis, and H. Mannila.

Selected publications

NeurIPS

Comparing Explanations is Not Enough, Explain the Change: New Standards are Needed to Explain Behavioral Shifts in Large Language Models

Martino Ciaperoni, Marzio Di Vece , Roberto Pellungrini , and 3 more authors

In Advances in Neural Information Processing Systems (NeurIPS), Position Paper Track , 2026

Accepted for publication

arXiv Bib

@inproceedings{ciaperoni2026comparing,
  title = {Comparing Explanations is Not Enough, Explain the Change: New Standards are Needed to Explain Behavioral Shifts in Large Language Models},
  author = {Ciaperoni, Martino and Di Vece, Marzio and Pellungrini, Roberto and Pappalardo, Luca and Giannotti, Fosca and Giannini, Francesco},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS), Position Paper Track},
  year = {2026},
  note = {Accepted for publication},
}

IJCAI

Efficient Optimization of Fixed-Length Paths

Martino Ciaperoni, Nikolaos Tziavelis , and Panagiotis Karras

In Proceedings of the 35th International Joint Conference on Artificial Intelligence (IJCAI) , 2026

DOI Bib

@inproceedings{ciaperoni2026efficient,
  title = {Efficient Optimization of Fixed-Length Paths},
  author = {Ciaperoni, Martino and Tziavelis, Nikolaos and Karras, Panagiotis},
  booktitle = {Proceedings of the 35th International Joint Conference on Artificial Intelligence (IJCAI)},
  pages = {2199--2207},
  year = {2026},
  doi = {10.24963/ijcai.2026/245},
}

KDD

Explainable AI for Cancer Drug Response Prediction: Beyond Univariate Feature Attributions

Martino Ciaperoni, M. Lalli , Simone Piaggesi , and 6 more authors

In Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) , 2026

DOI Bib

@inproceedings{ciaperoni2026explainable,
  title = {Explainable AI for Cancer Drug Response Prediction: Beyond Univariate Feature Attributions},
  author = {Ciaperoni, Martino and Lalli, M. and Piaggesi, Simone and Varisco, M. and Carli, F. and Guidotti, Riccardo and Pedreschi, Dino and Raimondi, Francesco and Giannotti, Fosca},
  booktitle = {Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD)},
  pages = {10749--10760},
  year = {2026},
  doi = {10.1145/3770855.3819010},
}

KDD

Efficient Exploration of the Rashomon Set of Rule-Set Models

Martino Ciaperoni, Han Xiao , and Aristides Gionis

In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) , 2024

DOI Bib

@inproceedings{ciaperoni2024efficient,
  title = {Efficient Exploration of the Rashomon Set of Rule-Set Models},
  author = {Ciaperoni, Martino and Xiao, Han and Gionis, Aristides},
  booktitle = {Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD)},
  pages = {478--489},
  year = {2024},
  doi = {10.1145/3637528.3671818},
}

DMKD

The Hadamard Decomposition Problem

Martino Ciaperoni, Aristides Gionis , and Heikki Mannila

Data Mining and Knowledge Discovery, 2024

DOI Bib Code

@article{ciaperoni2024hadamard,
  title = {The Hadamard Decomposition Problem},
  author = {Ciaperoni, Martino and Gionis, Aristides and Mannila, Heikki},
  journal = {Data Mining and Knowledge Discovery},
  volume = {38},
  number = {4},
  pages = {2306--2347},
  year = {2024},
  publisher = {Springer},
  doi = {10.1007/s10618-024-01033-y},
}

SIGMOD

SIEVE: A Space-Efficient Algorithm for Viterbi Decoding

Martino Ciaperoni, Aristides Gionis , Athanasios Katsamanis , and 1 more author

In Proceedings of the 2022 International Conference on Management of Data (SIGMOD) , 2022

Related patent application filed

DOI Bib

@inproceedings{ciaperoni2022sieve,
  title = {SIEVE: A Space-Efficient Algorithm for Viterbi Decoding},
  author = {Ciaperoni, Martino and Gionis, Aristides and Katsamanis, Athanasios and Karras, Panagiotis},
  booktitle = {Proceedings of the 2022 International Conference on Management of Data (SIGMOD)},
  pages = {1136--1145},
  year = {2022},
  doi = {10.1145/3514221.3526170},
  note = {Related patent application filed},
}