Research
Research focus #
My research focuses on Multimodal Large Language Models, with a current focus on SpeechLLMs, models that process spoken language. As a postdoctoral researcher at FBK, I have been working within the European projects Meetween, DVPS, and LLMs4EU. My PhD research was partially supported by the TANGO European Project, and during my research internship at Telefónica R&D in Barcelona I worked under the SPATIAL European Project on private and fair federated learning.
Publications #
Each paper has a short research post explaining the problem, approach, and main findings.
2026

Training Small LLMs from Scratch: Lessons from Zagreus and Nesso
CLiC-it 2026 · Palermo, Italy 🇮🇹

Interpreting Federated Learning by Aggregating SHAP Explanations
IEEE Access · Journal

Rashomon Sets and Model Multiplicity in Federated Learning
ACM Conference on Fairness, Accountability, and Transparency (FAccT) · Montréal, Canada 🇨🇦

FeDa4Fair: Client-Level Federated Datasets for Fairness Evaluation
ACM Conference on Fairness, Accountability, and Transparency (FAccT) · Montréal, Canada 🇨🇦
2025

An Interpretable Data-Driven Unsupervised Approach for the Prevention of Forgotten Items
28th European Conference on Artificial Intelligence (ECAI) · Bologna, Italy 🇮🇹

Enhancing Local Explanations with GAN-Based Neighborhood Generation
Discovery Science · Ljubljana, Slovenia 🇸🇮

A Benchmark for Client-Level Fairness in Federated Learning
Fourth European Workshop on Algorithmic Fairness (EWAF) · Eindhoven, The Netherlands 🇳🇱

Benefits of the Federation? Analyzing the Impact of Fair Federated Learning at the Client Level
ACM Conference on Fairness, Accountability, and Transparency (FAccT) · Athens, Greece 🇬🇷

Differentially Private FastSHAP for Federated Learning Model Explainability
International Joint Conference on Neural Networks (IJCNN) · Rome, Italy 🇮🇹
2024

PUFFLE: Balancing Privacy, Utility, and Fairness in Federated Learning
27th European Conference on Artificial Intelligence (ECAI) · Santiago de Compostela, Spain 🇪🇸

Enhancing Privacy and Utility in Federated Learning: A Hybrid P2P and Server-Based Approach with Differential Privacy Protection
21st International Conference on Security and Cryptography (SECRYPT) · Dijon, France 🇫🇷
2023

Semantic Enrichment of Explanations of AI Models for Healthcare
Discovery Science · Porto, Portugal 🇵🇹

Explaining Black-Boxes in Federated Learning
XAI World Conference · Lisbon, Portugal 🇵🇹
European projects #
These are European projects I have contributed to at different stages of my research.
Current contributions #
- Meetween: a Horizon Europe project developing multilingual meeting AI, including speech-to-speech translation, summarisation, and multimodal speech models. I am contributing to deliverables and to the fine-tuning of models.
- DVPS: a Horizon Europe project on multimodal foundation models combining speech, text, video, gestures, and other modalities. I am contributing to the development of Personalized Multimodal LLMs.
- LLMs4EU: an EU project focused on adapting and deploying language models across European languages and application areas.
Previous contributions #
- TANGO: a Horizon Europe project on hybrid human–machine decision making. My PhD research was partially supported by this project.
- SPATIAL — a Horizon 2020 project on secure, privacy-preserving, and trustworthy AI. During my research internship at Telefónica R&D in Barcelona, I worked on private and fair federated learning under SPATIAL.
📚 Thesis Co-Supervision #
- Elia Bisconti, “Towards Personalized Explainability: The development of a Modular Recommender System to Assess Explanation Suitability for Different Personas”, Master Degree in Computer Science, 2025
- Javier Alejandro Borges Legrottaglie, “A Data-Driven Unsupervised Approach for the Prevention of Forgotten Items", Master Degree in Digital Humanities, 2025
- Francesco Cappellini, “Federated Explainable Artificial Intelligence: A Case Study", One-year Master in Big Data Analytics & Social Mining, 2023
Reviewer for #
Conferences
- EMNLP 2026
- ECML-PKDD: 2024, 2025, 2026
- ECAI: 2024
- Discovery Science: 2024, 2026
Journals
- Computer Communications since February 2026
- TMLR since November 2025
✈️ Participation in conferences and Summer Schools #
- CLiC-it 2026: 14-16 September, 2026 - Palermo - Italy, 🇮🇹
- CISPA - ELLIS - Summer School 2025 04 August - 08 August, 2025 - Saarbrücken, Germany 🇩🇪
- IJCNN 2025 - 30 June - 5 July, 2025- Rome, Italy 🇮🇹
- Facct 2025 - 22 June - 26 June, 2025 - Athens, Greece 🇬🇷
- ECAI 2024 - 21 October - 24 October, 2024 - Santiago De Compostela, Spain 🇪🇸
- Summer School on Security & Privacy in the Age of AI 2024 - 10 September - 13 September, 2024 - Leuven, Belgium 🇧🇪
- M2L: Mediterranean Machine Learning Summer School - 28th August - 2nd September, 2023 - Thessaloniki, Greece 🇬🇷
- XAI World Conference 2023 - 25th - July 28th, 2023. Lisbon, Portugal 🇵🇹
- Lipari Summer School - DATA SCIENCE: Models, Algorithms, AI and Beyond July - 17th - July 23rd, 2022. Lipari 🇮🇹
- HHAI: The first International Conference on Hybrid Human-Artificial Intelligence - 13-17 June 2022, Amsterdam 🇳🇱