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 Luxembourg Ministry for Digitalisation, within the framework of the Tech-in-GOV initiative, 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.
As Large Language Models are deployed globally, claims of “cultural awareness” often mask homogenised outputs. This workshop bridges computational linguistics, critical AI evaluation, and regional sociolinguistics to empirically investigate “cultural awareness” and “cultural reasoning”. We focus on the Greater Region of SaarLorLux, which is a uniquely multilingual, cross-border landscape spanning Luxembourg, as well as areas of Germany, France, and Belgium, where overlapping identities, institutional practices, and language varieties provide a real-world testbed for AI cultural adaptability.
To move beyond the more abstract “cultural awareness”, this year’s workshop partners with the EVAS Research Cooperative (Ecological Validation of Artificial Simulacra), whose mission is to systematically test whether language models can legitimately serve as behavioral or cultural proxies rather than just assuming they can. Together with EVAS’s empirical validation framework, we present and discuss the ongoing CLUG (Cultural Language Understanding in the Greater Region) project: a multilingual dataset with cultural annotation. Together, these initiatives will serve as exemples for whether culturally embedded meaning survives or collapses during machine translation and generative downstream tasks, a topic we wish to discuss more broadly at this workshop.
We warmly invite researchers from all disciplinary and technical backgrounds—including linguistics, computer science, sociology, anthropology, philosophy, and cognitive science. No prior expertise in NLP or AI evaluation is required; what matters is a commitment to empirically grounded, cross-disciplinary inquiry into how culture, language, and technology intersect within AI.
Please note that participation is free, but registration is mandatory (see Keydates and Registration Info)
To frame our discussions, we plan to host keynote speakers specializing in cultural NLP, critical AI evaluation, and regional sociolinguistics. These presentations will set the stage for collaborative validation exercises, paper discussions, and open dialogue. We invite papers/abstracts around conceptual framing, as well as testable, ecologically valid LLM analysis. Submissions may focus on, but are not limited to:
This workshop serves as the launchpad for a dedicated Special Issue focused on empirical validation and cross-border language processing. We welcome submissions of abstracts or full papers that will be presented as posters or talks at the event. Accepted full paper workshop contributions may be invited to submit manuscripts for peer review, with selected papers published in the associated Special Issue following successful evaluation (as well as acceptance of the Special Issue).
Submit abstracts/papers to: [link to follow (enquiries: alistair.plum@uni.lu)] </br> Deadline: 13 October 2026 (AOE) </br> Format: We accept abstracs, as well as papers up to 8 pages, following the ACL formatting guidelines. Submission will be via openreview.net </br> Acceptance Notification: 27 October 2026 (AOE) </br> Note: First acceptance grants presentation rights at the workshop. Final publication decisions for the Special Issue will be made post-workshop in consultation with the guest editors and journal publishers.
Workshop: 18 November 2026, 10:00–17:00 </br> Please Sign-up here: https://forms.gle/VjcYnifYVUeZRqRA9 </br> Note: Participation is free and not contingent on handing in an abstract or paper, but prior registration is mandatory in all cases.
Black Box, Ground Floor, Maison du Sciences Humaines University of Luxembourg, Belval Campus
Organisers:
Project Links: </br> CLUG: https://curia-gr.github.io </br> EVAS: https://evas-research.github.io
With the support of the Luxembourg Ministry for Digitalisation, within the framework of the Tech-in-GOV initiative.