2026

The Right to Hide: Masking Community Affiliation via Minimal Graph Rewiring

OSNEM

Matteo Silvestri, Edoardo Gabrielli, Fabrizio Silvestri, Gabriele Tolomei

Online Social Networks and Media · 55

Abstract

Protecting privacy in social graphs may require obscuring nodes' membership in sensitive communities. However, doing so without significantly disrupting the underlying graph topology remains a key challenge. In this work, we address the community membership hiding problem, which involves strategically modifying the graph structure to conceal a target node's affiliation with a community, regardless of the detection algorithm used. We reformulate the original discrete, counterfactual graph search objective as a differentiable constrained optimisation task. To this end, we introduce ∇-CMH, a new gradient-based method that operates within a feasible modification budget to minimise structural changes while effectively hiding a node's community membership. Extensive experiments on multiple datasets and community detection methods demonstrate that our technique outperforms existing baselines, achieving the best balance between node hiding effectiveness and graph rewiring cost, while preserving computational efficiency.

Cite / BibTeX
@article{silvestri2026righthidemaskingcommunity,
 abbr = {OSNEM},
 abstract = {Protecting privacy in social graphs may require obscuring nodes' membership in sensitive communities. However, doing so without significantly disrupting the underlying graph topology remains a key challenge. In this work, we address the community membership hiding problem, which involves strategically modifying the graph structure to conceal a target node's affiliation with a community, regardless of the detection algorithm used. We reformulate the original discrete, counterfactual graph search objective as a differentiable constrained optimisation task. To this end, we introduce ∇-CMH, a new gradient-based method that operates within a feasible modification budget to minimise structural changes while effectively hiding a node's community membership. Extensive experiments on multiple datasets and community detection methods demonstrate that our technique outperforms existing baselines, achieving the best balance between node hiding effectiveness and graph rewiring cost, while preserving computational efficiency.},
 archiveprefix = {arXiv},
 author = {Silvestri, Matteo and Gabrielli, Edoardo and Silvestri, Fabrizio and Tolomei, Gabriele},
 code = {https://github.com/hercolelab/community-membership-hiding},
 doi = {10.1016/j.osnem.2026.100357},
 eprint = {2502.00432},
 journal = {Online Social Networks and Media},
 month = {Sep},
 pages = {100357},
 pdf = {https://www.sciencedirect.com/science/article/pii/S2468696426000133},
 primaryclass = {cs.SI},
 publisher = {Elsevier},
 title = {The Right to Hide: Masking Community Affiliation via Minimal Graph Rewiring},
 url = {https://www.sciencedirect.com/science/article/pii/S2468696426000133},
 volume = {55},
 year = {2026}
}

Explaining AI Agents Through Execution Traces

Preprint

Vittoria Vineis, Fabiano Veglianti, Lorenzo Antonelli, Claudia Di Carlo, Matteo Silvestri, Gabriele Tolomei

Abstract

AI Agents are increasingly deployed in real-world settings, where they interact with external tools and make sequential decisions with limited human oversight. This creates a pressing need for reliable and auditable explanations of what an agent did and why. However, traditional Explainable AI (XAI) methods fall short of providing the process-level transparency required for such interactive, multi-step systems, motivating a paradigm shift toward approaches specifically designed for AI Agents. To address this gap, we present a post-hoc XAI framework that transforms a lengthy agent's execution trace into a structured report and a faithful natural-language explanation explicitly grounded in its observable behavior. Because it relies solely on execution traces, the framework applies across different agent architectures, environments, and tasks. Human and automated evaluations across multiple benchmarks and architectures show that our framework produces high-quality, trace-faithful explanations while reliably identifying unsupported claims, unjustified actions, and evidence gaps, outperforming naive LLM-generated explanations.

Cite / BibTeX
@misc{vineis2026explainingagents,
 abbr = {Preprint},
 abstract = {AI Agents are increasingly deployed in real-world settings, where they interact with external tools and make sequential decisions with limited human oversight. This creates a pressing need for reliable and auditable explanations of what an agent did and why. However, traditional Explainable AI (XAI) methods fall short of providing the process-level transparency required for such interactive, multi-step systems, motivating a paradigm shift toward approaches specifically designed for AI Agents. To address this gap, we present a post-hoc XAI framework that transforms a lengthy agent's execution trace into a structured report and a faithful natural-language explanation explicitly grounded in its observable behavior. Because it relies solely on execution traces, the framework applies across different agent architectures, environments, and tasks. Human and automated evaluations across multiple benchmarks and architectures show that our framework produces high-quality, trace-faithful explanations while reliably identifying unsupported claims, unjustified actions, and evidence gaps, outperforming naive LLM-generated explanations.},
 archiveprefix = {arXiv},
 author = {Vineis, Vittoria and Veglianti, Fabiano and Antonelli, Lorenzo and Di Carlo, Claudia and Silvestri, Matteo and Tolomei, Gabriele},
 eprint = {2609.06063},
 month = {Sep},
 pdf = {https://arxiv.org/pdf/2609.06063},
 primaryclass = {cs.AI},
 title = {Explaining AI Agents Through Execution Traces},
 url = {https://arxiv.org/abs/2609.06063},
 year = {2026}
}

EMBL AI Librarian: Life-Sciences Knowledge Layer for AI Agents

Preprint

Luigi Sigillo, Matteo Silvestri, Francesco Tabaro, Rajat Bhatnagar, Syed Irtaza Mubashar, Matt Jeffryes, Daljit Nijjer, Vittorio Perera, Ola Spjuth, Julio Saez-Rodriguez, Melissa Harrison, Fabio Petroni

Abstract

The web is increasingly accessed by AI agents rather than humans. Every agent needs knowledge, especially in the life-sciences, where agentic pipelines are growing fast. Access to the literature is a crucial part of that need, and resources such as Europe PMC, with over 40M indexed records, are widely used to meet it. Yet these resources were not built for AI agents: they take keywords and complex syntax and return whole papers, so every agent must learn the syntax, issue several searches, and read full papers to find the evidence it needs. We introduce EMBL AI Librarian, a knowledge layer that upgrades the Europe PMC interface for AI agents: an agent asks in natural language and receives evidence that answers it. A single LLM orchestrates the whole knowledge retrieval process: it plans complementary subqueries executed by the live Europe PMC search engine, then reads the selected papers and locates the relevant evidence. We evaluate Librarian across four benchmarks: literature synthesis, claim verification, open-domain question answering, and downstream biology tasks such as protocol questions and sequence manipulation. On ScholarQABench, Librarian improves Citation F1 by more than $16$ points over strong recently published baselines. Used as the retrieval layer of an existing claim-verification pipeline, it increases agreement with expert consensus; and on the open-form LitQA2 benchmark, a GPT-5.4 agent scores about $8$ points higher when grounded in Librarian than with web search. Overall, our results show that equipping life-science agents with the Librarian knowledge layer improves performance across a range of tasks. We release our code publicly at https://github.com/petroni-lab/librarian.

Cite / BibTeX
@misc{sigillo2026emblailibrarian,
 abbr = {Preprint},
 abstract = {The web is increasingly accessed by AI agents rather than humans. Every agent needs knowledge, especially in the life-sciences, where agentic pipelines are growing fast. Access to the literature is a crucial part of that need, and resources such as Europe PMC, with over 40M indexed records, are widely used to meet it. Yet these resources were not built for AI agents: they take keywords and complex syntax and return whole papers, so every agent must learn the syntax, issue several searches, and read full papers to find the evidence it needs. We introduce EMBL AI Librarian, a knowledge layer that upgrades the Europe PMC interface for AI agents: an agent asks in natural language and receives evidence that answers it. A single LLM orchestrates the whole knowledge retrieval process: it plans complementary subqueries executed by the live Europe PMC search engine, then reads the selected papers and locates the relevant evidence. We evaluate Librarian across four benchmarks: literature synthesis, claim verification, open-domain question answering, and downstream biology tasks such as protocol questions and sequence manipulation. On ScholarQABench, Librarian improves Citation F1 by more than $16$ points over strong recently published baselines. Used as the retrieval layer of an existing claim-verification pipeline, it increases agreement with expert consensus; and on the open-form LitQA2 benchmark, a GPT-5.4 agent scores about $8$ points higher when grounded in Librarian than with web search. Overall, our results show that equipping life-science agents with the Librarian knowledge layer improves performance across a range of tasks. We release our code publicly at https://github.com/petroni-lab/librarian.},
 archiveprefix = {arXiv},
 author = {Sigillo, Luigi and Silvestri, Matteo and Tabaro, Francesco and Bhatnagar, Rajat and Mubashar, Syed Irtaza and Jeffryes, Matt and Nijjer, Daljit and Perera, Vittorio and Spjuth, Ola and Saez-Rodriguez, Julio and Harrison, Melissa and Petroni, Fabio},
 code = {https://github.com/petroni-lab/librarian},
 eprint = {2607.28229},
 month = {Jul},
 pdf = {https://arxiv.org/pdf/2607.28229},
 primaryclass = {cs.CL},
 title = {EMBL AI Librarian: Life-Sciences Knowledge Layer for AI Agents},
 url = {https://arxiv.org/abs/2607.28229},
 year = {2026}
}

When Large Language Models Know the Table: A Framework for Assessing Data Contamination in Tabular Datasets

COLM '26

Matteo Silvestri, Fabiano Veglianti, Flavio Giorgi, Fabrizio Silvestri, Gabriele Tolomei

Abstract

Large language models (LLMs) are increasingly exposed to data contamination, i.e., performance gains driven by prior exposure of test datasets rather than generalization. However, in the context of tabular data, this problem is largely unexplored. Existing approaches primarily rely on memorization tests, which are too coarse to detect contamination. In contrast, we propose a framework for assessing contamination in tabular datasets by generating controlled queries and performing comparative evaluation. Given a dataset, we craft multiple-choice aligned queries that preserve task structure while allowing systematic transformations of the underlying data. These transformations are designed to selectively disrupt dataset information while preserving partial knowledge, enabling us to isolate performance attributable to contamination. We complement this setup with non-neural baselines that provide reference performance, and we introduce a statistical testing procedure to formally detect significant deviations indicative of contamination. Empirical results on eight widely used tabular datasets reveal clear evidence of contamination in four cases. These findings suggest that performance on downstream tasks involving such datasets may be substantially inflated, raising concerns about the reliability of current evaluation practices.

Cite / BibTeX
@article{silvestri2025evaluating,
 abbr = {COLM '26},
 abstract = {Large language models (LLMs) are increasingly exposed to data contamination, i.e., performance gains driven by prior exposure of test datasets rather than generalization. However, in the context of tabular data, this problem is largely unexplored. Existing approaches primarily rely on memorization tests, which are too coarse to detect contamination. In contrast, we propose a framework for assessing contamination in tabular datasets by generating controlled queries and performing comparative evaluation. Given a dataset, we craft multiple-choice aligned queries that preserve task structure while allowing systematic transformations of the underlying data. These transformations are designed to selectively disrupt dataset information while preserving partial knowledge, enabling us to isolate performance attributable to contamination. We complement this setup with non-neural baselines that provide reference performance, and we introduce a statistical testing procedure to formally detect significant deviations indicative of contamination. Empirical results on eight widely used tabular datasets reveal clear evidence of contamination in four cases. These findings suggest that performance on downstream tasks involving such datasets may be substantially inflated, raising concerns about the reliability of current evaluation practices.},
 archiveprefix = {arXiv},
 author = {Silvestri, Matteo and Veglianti, Fabiano and Giorgi, Flavio and Silvestri, Fabrizio and Tolomei, Gabriele},
 code = {https://github.com/hercolelab/tabular_contamination},
 eprint = {2510.20351},
 month = {Mar},
 pdf = {https://arxiv.org/pdf/2510.20351},
 primaryclass = {cs.LG},
 title = {When Large Language Models Know the Table: A Framework for Assessing Data Contamination in Tabular Datasets},
 url = {https://arxiv.org/abs/2510.20351},
 year = {2026}
}

PONTE: Personalized Orchestration for Natural Language Trustworthy Explanations

XAI '26

Vittoria Vineis, Matteo Silvestri, Lorenzo Antonelli, Filippo Betello, Gabriele Tolomei

Abstract

Explainable Artificial Intelligence (XAI) seeks to enhance the transparency and accountability of machine learning systems, yet most methods follow a one-size-fits-all paradigm that neglects user differences in expertise, goals, and cognitive needs. Although Large Language Models can translate technical explanations into natural language, they introduce challenges related to faithfulness and hallucinations. To address these challenges, we present PONTE (Personalized Orchestration for Natural language Trustworthy Explanations), a human-in-the-loop framework for adaptive and reliable XAI narratives. PONTE models personalization as a closed-loop validation and adaptation process rather than prompt engineering. It combines: (i) a low-dimensional preference model capturing stylistic requirements; (ii) a preference-conditioned generator grounded in structured XAI artifacts; and (iii) verification modules enforcing numerical faithfulness, informational completeness, and stylistic alignment, optionally supported by retrieval-grounded argumentation. User feedback iteratively updates the preference state, enabling quick personalization. Automatic and human evaluations across healthcare and finance domains show that the verification–refinement loop substantially improves completeness and stylistic alignment over validation-free generation. Human studies further confirm strong agreement between intended preference vectors and perceived style, robustness to generation stochasticity, and consistently positive quality assessments.

Cite / BibTeX
@article{vineis2026ponte,
 abbr = {XAI '26},
 abstract = {Explainable Artificial Intelligence (XAI) seeks to enhance the transparency and accountability of machine learning systems, yet most methods follow a one-size-fits-all paradigm that neglects user differences in expertise, goals, and cognitive needs. Although Large Language Models can translate technical explanations into natural language, they introduce challenges related to faithfulness and hallucinations. To address these challenges, we present PONTE (Personalized Orchestration for Natural language Trustworthy Explanations), a human-in-the-loop framework for adaptive and reliable XAI narratives. PONTE models personalization as a closed-loop validation and adaptation process rather than prompt engineering. It combines: (i) a low-dimensional preference model capturing stylistic requirements; (ii) a preference-conditioned generator grounded in structured XAI artifacts; and (iii) verification modules enforcing numerical faithfulness, informational completeness, and stylistic alignment, optionally supported by retrieval-grounded argumentation. User feedback iteratively updates the preference state, enabling quick personalization. Automatic and human evaluations across healthcare and finance domains show that the verification–refinement loop substantially improves completeness and stylistic alignment over validation-free generation. Human studies further confirm strong agreement between intended preference vectors and perceived style, robustness to generation stochasticity, and consistently positive quality assessments.},
 archiveprefix = {arXiv},
 author = {Vineis, Vittoria and Silvestri, Matteo and Antonelli, Lorenzo and Betello, Filippo and Tolomei, Gabriele},
 eprint = {2603.06485},
 month = {Mar},
 pdf = {https://arxiv.org/pdf/2603.06485},
 primaryclass = {cs.LG},
 title = {PONTE: Personalized Orchestration for Natural Language Trustworthy Explanations},
 url = {https://arxiv.org/abs/2603.06485},
 year = {2026}
}

Exploring the potential of artificial intelligence in assessing the risk of gastric neoplastic lesions in patients with corpus atrophic gastritis

Springer Nature

Emanuele Dilaghi, Edoardo Cesaroni, Irene Ligato, Matteo Silvestri, Giampaolo Liuzzi, Bruno Annibale, Stefano Lucidi, Gianluca Esposito, Marco Sciandrone

Gastric Cancer · 29

Abstract

Corpus atrophic gastritis (CAG) requires endoscopic-histological surveillance due to the risk of developing gastric neoplastic lesions (GNL). This study aimed to identify variables associated with GNL development at long-term follow-up using a Fisher score-based feature-ranking-approach coupled with a One-Class Support-Vector-Machine (SVM) model.

Cite / BibTeX
@article{Dilaghi2025,
 abbr = {Springer Nature},
 abstract = {Corpus atrophic gastritis (CAG) requires endoscopic-histological surveillance due to the risk of developing gastric neoplastic lesions (GNL). This study aimed to identify variables associated with GNL development at long-term follow-up using a Fisher score-based feature-ranking-approach coupled with a One-Class Support-Vector-Machine (SVM) model.},
 author = {Dilaghi, Emanuele
and Cesaroni, Edoardo
and Ligato, Irene
and Silvestri, Matteo
and Liuzzi, Giampaolo
and Annibale, Bruno
and Lucidi, Stefano
and Esposito, Gianluca
and Sciandrone, Marco},
 doi = {10.1007/s10120-025-01679-7},
 issn = {1436-3305},
 journal = {Gastric Cancer},
 number = {1},
 pages = {159--168},
 pdf = {https://doi.org/10.1007/s10120-025-01679-7},
 title = {Exploring the potential of artificial intelligence in assessing the risk of gastric neoplastic lesions in patients with corpus atrophic gastritis},
 url = {https://doi.org/10.1007/s10120-025-01679-7},
 volume = {29},
 year = {2026}
}

2025

A Survey on Explainable AI Narratives based on Large Language Models

Preprint

Matteo Silvestri, Vittoria Vineis, Edoardo Gabrielli, Flavio Giorgi, Fabiano Veglianti, Fabrizio Silvestri, Gabriele Tolomei

TechRxiv

Abstract

Explainable Artificial Intelligence (XAI) seeks to elucidate the inner logic of machine learning models, yet its outputs often remain difficult for non-technical users to understand. The emerging paradigm of XAI Narratives leverages Large Language Models (LLMs) to translate technical explanations into coherent, human-readable accounts. This survey provides the first systematic review of this approach, focusing on systems in which LLMs act as post-hoc narrative translators rather than autonomous explainers. We formalize this task as the Narrative Generation Problem, examine its integration with classical XAI methods such as feature attribution and counterfactual explanations across multiple data modalities, and introduce a taxonomy for narrative evaluation spanning three core dimensions. Finally, we analyze prompting strategies and outline open challenges and future directions for advancing reliable, interpretable, and context-aware XAI Narratives.

Cite / BibTeX
@article{silvestri2025survey,
 abbr = {Preprint},
 abstract = {Explainable Artificial Intelligence (XAI) seeks to elucidate the inner logic of machine learning models, yet its outputs often remain difficult for non-technical users to understand. The emerging paradigm of XAI Narratives leverages Large Language Models (LLMs) to translate technical explanations into coherent, human-readable accounts. This survey provides the first systematic review of this approach, focusing on systems in which LLMs act as post-hoc narrative translators rather than autonomous explainers. We formalize this task as the Narrative Generation Problem, examine its integration with classical XAI methods such as feature attribution and counterfactual explanations across multiple data modalities, and introduce a taxonomy for narrative evaluation spanning three core dimensions. Finally, we analyze prompting strategies and outline open challenges and future directions for advancing reliable, interpretable, and context-aware XAI Narratives.},
 author = {Silvestri, Matteo and Vineis, Vittoria and Gabrielli, Edoardo and Giorgi, Flavio and Veglianti, Fabiano and Silvestri, Fabrizio and Tolomei, Gabriele},
 journal = {TechRxiv},
 month = {Nov},
 pdf = {https://www.techrxiv.org/users/995230/articles/1356253-a-survey-on-explainable-ai-narratives-based-on-large-language-models},
 title = {A Survey on Explainable AI Narratives based on Large Language Models},
 url = {https://www.techrxiv.org/users/995230/articles/1356253-a-survey-on-explainable-ai-narratives-based-on-large-language-models},
 year = {2025}
}

Enhancing XAI Narratives through Multi-Narrative Refinement and Knowledge Distillation

HCAI | CIKM '25

Flavio Giorgi, Matteo Silvestri, Cesare Campagnano, Fabrizio Silvestri, Gabriele Tolomei

Abstract

Explainable Artificial Intelligence has become a crucial area of research, aiming to demystify the decision-making processes of deep learning models. Among various explainability techniques, counterfactual explanations have been proven particularly promising, as they offer insights into model behavior by highlighting minimal changes that would alter a prediction. Despite their potential, these explanations are often complex and technical, making them difficult for non-experts to interpret. To address this challenge, we propose a novel pipeline that leverages Language Models, large and small, to compose narratives for counterfactual explanations. We employ knowledge distillation techniques along with a refining mechanism to enable Small Language Models to perform comparably to their larger counterparts while maintaining robust reasoning abilities. In addition, we introduce a simple but effective evaluation method to assess natural language narratives, designed to verify whether the models' responses are in line with the factual, counterfactual ground truth. As a result, our proposed pipeline enhances both the reasoning capabilities and practical performance of student models, making them more suitable for real-world use cases.

Cite / BibTeX
@misc{giorgi2025enhancingxainarrativesmultinarrative,
 abbr = {HCAI | CIKM '25},
 abstract = {Explainable Artificial Intelligence has become a crucial area of research, aiming to demystify the decision-making processes of deep learning models. Among various explainability techniques, counterfactual explanations have been proven particularly promising, as they offer insights into model behavior by highlighting minimal changes that would alter a prediction. Despite their potential, these explanations are often complex and technical, making them difficult for non-experts to interpret. To address this challenge, we propose a novel pipeline that leverages Language Models, large and small, to compose narratives for counterfactual explanations. We employ knowledge distillation techniques along with a refining mechanism to enable Small Language Models to perform comparably to their larger counterparts while maintaining robust reasoning abilities. In addition, we introduce a simple but effective evaluation method to assess natural language narratives, designed to verify whether the models' responses are in line with the factual, counterfactual ground truth. As a result, our proposed pipeline enhances both the reasoning capabilities and practical performance of student models, making them more suitable for real-world use cases.},
 archiveprefix = {arXiv},
 author = {Flavio Giorgi and Matteo Silvestri and Cesare Campagnano and Fabrizio Silvestri and Gabriele Tolomei},
 code = {https://github.com/flaat/llm_kd},
 eprint = {2510.03134},
 month = {Sep},
 pdf = {https://arxiv.org/pdf/2510.03134},
 primaryclass = {cs.LG},
 title = {Enhancing XAI Narratives through Multi-Narrative Refinement and Knowledge Distillation},
 url = {https://arxiv.org/abs/2510.03134},
 year = {2025}
}