Serval Research Group

The Clarification Challenge Competition, SDKs, implementations of baseline algorithms, and public benchmarks.

This is the repository for AI security labs given at AI and Cybersecurity course for the Erasmus Mundus Joint Master in Cybersecurity, for the academic year 2026-2027.

A benchmark of XAI methods for VLMs models using faithfulness metrics (including a novel faithfulness metric to measure cross-modal reasoning in VLMs).

CERTAIN EU (WP5) -- Financial Asset Recommendation with counterfactual explanations

Membership Inference Attacks on Tabular Foundations Models

Can Large Language Models Reason and Optimize Under Constraints?