As part of the Saarland Informatics Campus, the RTG Neuroexplicit Models will contribute to the German Unity Day celebrations in Saarbrücken with three interactive demos. Visitors are invited to explore our research on explainable and transparent AI and to discuss current challenges with our researchers.

Our program:

🔹 B-cos LM (presented by Yifan Wang)
We will showcase a novel type of neural network that provides faithful and human-interpretable explanations for its predictions. Visitors can try out the model on sentiment analysis and toxicity detection tasks and examine the rationale behind each prediction
📅 Thursday, October 2 | ⏰ 15:00–16:30

🔹 Robustness of large language models in moral judgements (presented by Soyoung Oh)
This demo highlights inconsistencies in moral judgments of large language models: tiny changes in wording or labeling (e.g., “Case 1/Case 2” vs. “A/B”) can lead to completely different outcomes.
📅 Thursday, October 2 | ⏰ 16:30–17:00

🔹 SLayR: Steuerbare Bildgenerierung (presented by Cameron Braunstein)
Our SLayR system investigates how AI can generate plausible images from natural language input. By integrating user-defined layouts, it enables more realistic results than standard models – for example, avoiding TVs floating above fireplaces.
📅 Friday, October 3 | ⏰ 10:00–12:00

🔻 Location: Mainzerstraße 3, Saarbrücken city center (Google maps)

We warmly invite everyone to stop by, try out our demos, and exchange ideas with our researchers. Join us to experience how AI can be made more explainable, robust, and trustworthy.