Devana×LIST

Devana × LIST
Scientific Partnership

Advancing explainable AI research and innovation
for Luxembourg and Europe

Partnership overview

Since 2021, Devana has developed explainable AI systems for critical and regulated environments. This partnership with the Luxembourg Institute of Science and Technology (LIST) and the University of Luxembourg, formalized in 2025, extends this work and brings independent scientific validation to it.

LIST, Luxembourg's leading public applied research institute, brings scientific expertise in AI and validation of evaluation protocols. The University of Luxembourg supervises the doctoral training and awards the degree.

Together, we are advancing the state of the art in Explainable AI Systems in production contexts (X·AIS).

APPLIED RESEARCH

Key focus areas

Explainable AI research
A four-year doctoral thesis co-supervised by Devana, LIST, and the University of Luxembourg: "Explainable AI for Generative Models in RAG Contexts". Methods studied: mechanistic interpretability, source attribution, reference resolution.
Evaluation and benchmarks
Design and validation of multidimensional evaluation protocols to measure quality, source fidelity, and reliability of LLM-generated responses in RAG contexts. Identification and quantification of hallucinations, biases, and failures.
From research to production
Scientifically validated protocols are integrated into Devana to support decision-making in critical domains: healthcare, finance, public safety. Partnership outputs are intended for publication under recognized open-source licenses.
European Trustworthy AI standards
The work contributes to emerging Explainable AI standards and shapes evaluation practices for industry and academia. An approach aligned with the EU AI Act principles: transparency, robustness, fairness, accountability.
Current projects

RAG Score™ framework

Joint development of the RAG Score™ methodology to measure and improve the reliability of retrieval-augmented generation (RAG) systems. Work on source traceability, hallucination detection, and confidence calibration.

AI governance standards

Collaborative work on establishing European standards for AI governance, auditability, and compliance frameworks aligned with the EU AI Act.

LIST AI Sandbox

Access to the LIST AI Sandbox to experiment and validate evaluation methods in a dedicated research environment. Leveraging research technologies developed by LIST: FAIRGAME, LLM Factory, MLA-BiTe, KG-RAG.

Impact & outcomes
3
Publications for 2026
6
Researchers involved
3
Partner institutions

LEADERSHIP

Partnership team

LIST partnership team

German Castignani, Tom Lucas, and Alessio Buscemi at LIST, September 2024

About LIST

The Luxembourg Institute of Science and Technology (LIST) is a public applied research institute (Research and Technology Organisation, RTO) that develops market-oriented technologies for Luxembourg's economic sectors.

Our collaboration takes place within the AIRA Research Group (AI-Readiness and Assessment), specialized in AI systems and their evaluation.

Visit LIST Website →