curia-gr

The CLUG Project

The CLUG Project (Cultural Language Understanding Evaluation for the Greater Region) develops multilingual, pluricultural benchmarks to ensure that emerging AI technologies are sensitive to the specific linguistic and cultural contexts of the SaarLorLux region.

By enriching the ltzGLUE benchmark with cultural annotations—including idioms, historical facts, and cross-border entities—and providing translations across Luxembourgish, French, German, and English, CLUG provides the essential tools for public administrations to implement inclusive, context-aware AI solutions like chatbots and document classifiers.

With the support of the Tech-in-GOV initiative under from the Luxembourg Ministry for Digitalisation, our mission is to prevent the “flattening” of cultural nuance in machine learning models and to promote robust, ecologically valid language technologies for the Greater Region.

---

2nd Workshop on Artificial Cultural Reasoning: Empirical Validation & Cross-Border Language

Theme: Moving beyond cultural awareness to testable, ecologically valid AI behavior.

This workshop builds on our previous conceptual discussions by shifting toward empirical validation. In collaboration with the EVAS Research Cooperative and the CLUG Project Team, we will stress-test whether current AI systems can genuinely handle cross-border cultural reasoning—or whether they merely reproduce dominant perspectives under the guise of fluency.

The Validation Pipeline

This year’s workshop utilizes a concrete pipeline to move past abstract claims through three core pillars:

1. CLUG’s 4-Dimensional Rubric: We use a standardized evaluation metric covering:
   • Cultural Reference: (Institutional, historical, and geographical context)
   • Directionality: (Inward vs. outward framing)
   • Perspective & Anchor: (DE/FR/LU vantage points)
   • Cultural Register: (Local discourse conventions, formality, and hedging)
2. Translation Scoring: Measuring the degree to which cultural nuance is lost or “flattened” during cross-lingual generation.
3. EVAS’s Critical Lens: Testing whether surface-level fluency in LLMs masks a failure to respect local cultural norms in border regions.

Organisers

• Dr. Alistair Plum (Université du Luxembourg)
• Dr. Simon Münker (Universität Trier)
• Prof. Dr. Achim Rettinger (Universität Trier)

In collaboration with the EVAS Research Cooperative and the CLUG Project Team

---

Abstract Submission

We invite submissions that align with our validation focus. Accepted contributions will be presented during the poster or session blocks.

Suggested Topics:

• Applying CLUG’s 4 dimensions to LLM evaluation.
• Cross-lingual preservation of cultural register or pragmatic norms.
• Critical analysis of “cultural neutrality” vs. adaptive AI in border regions.
• Open datasets, fine-tuning strategies, or prompt-engineering for cultural reasoning.
• Methodological reflections on inter-annotator agreement and ecological validity.

Submission Guidelines:

• TBD

---

Important Information

Dates

• Workshop: 18th November 2026 (tentative)
• Registration Deadline: TBD
• Abstracts Due: TBD

Location

Black Box Maison du Sciences Humaines
Université du Luxembourg

Registration

Attendance is free. Please register via the link below:
TBD

Contact

alistair.plum@uni.lu

---