relAI at TUM Maus Tag 2026

mc

AI, curiosity, and lots of smart questions at Maus-Tag on the TUM Campus!

At Maus-Tag at TUM Campus, relAI had the pleasure of showing young visitors that AI is not just something to use, but something to question, explore, and understand.

Guided by relAI PhD student Max Beier, and relAI coordinator Anja Woracek, first thing our young visitors learned about was gradient descent. They explored this key machine learning concept through the KI-Campus "Ship & Treasure" game, discovering how AI learns step by step to find the best solution, just like navigating toward a hidden treasure.

Then, the kids trained their own image classification models with fruits. They recorded images, trained the model, and then immediately started testing its limits: What happens if we show a different object? Why does the model get it right? Why does it get it wrong? The list of questions was long!

What impressed us most was the kids scientific mindset. They didn't simply trust the AI's predictions. They experimented and challenged the system. Real little researchers and critical thinkers.

Want to try it yourself?

🍎 Teachable Machine: https://teachablemachine.withgoogle.com/
⚓ Gradient Descent Game: https://ki-campus.org/simulations/gradient-descent

We are deeply grateful to the organizers, especially Dr. Isabella Almstätter from MDSI, whose support made the day possible. A big thank you to  relAI student Max Beier for sparking curiosity and plenty of "aha!" Moments among the kids throughout the day.

Most of all, a huge shout-out to all the curious kids who joined us on this AI adventure. Your questions, creativity, and willingness to challenge the models reminded us what great research is all about. The future of AI is in good hands.

👉 Visit the MDSI website for more details and picture gallery: From molecules to machines: A hands-on journey through science - Munich Data Science Institute 

On November 9, 2026, starting at 6:15 PM, the annual LMU AI Symposium will be held in the Great Hall (Große Aula) of LMU Munich. This event is organized by AI-HUB@LMU in collaboration with relAI and will be followed by a reception in the Lichthof. We anticipate welcoming approximately 500 guests from academia, industry, politics, and society.

As in previous years, a central highlight will be the pitch talks by our newly appointed AI professors, showcasing the breadth and diversity of AI research at LMU.

Further highlights of the program include:

  • A keynote by relAI Fellow Prof. Dr. Stefan Feuerriegel on “Scientific Discovery in the AI Age”
  • A panel discussion on “Prompt or Perish? Generative AI and the University Teaching of Tomorrow”
  • A talk by LMU Vice President Dr. Philipp Baaske entitled “IP Transfer@LMU: worry-free and fast”
  • The AI-HUB@LMU Awards Ceremony, supported by the Münchener Universitätsgesellschaft

You can find the full program and registration details here: KI-Symposium 2026 - AI-HUB@LMU - LMU München

We would be delighted to welcome many of you to this event! 

As in the previous edition, relAI was present this year at the TUM Open Campus Day at Forschungscampus Garching on 3rd October 2026.

Throughout the day, a lively mix of visitors—families, students, people of all ages—stopped by our stand, eager to discover what relAI is all about. Guests got hands-on with AI, training their own neural networks and exploring intriguing concepts like Simpson's Paradox and the secrets behind how AI makes decisions.

We are deeply grateful to the organizers, especially Dr. Isabella Almstätter from MDSI and Dr. Gihanna Galindez, whose support made the day possible. A big thank you to the relAI Coordinators, to relAI Student Muhammad ElMahdy for his enthusiasm, and to relAI Students Sameer Ambekar, Julian Balletshofer, Sarah Lumpp, Amine Ketata, Maria Matveev, Daniela Schkoda, and Richard Schwank for providing  the games for the event. Most of all, we appreciate everyone who stopped by out of curiosity and joined us on the adventure of exploring AI.

👉 Visit the MDSI website for more details and picture gallery: https://go.tum.de/394618

 

🐘The old story of the blind men and the elephant feels surprisingly relevant to medical AI.

One blind man touches the trunk and thinks the elephant is a snake. Another feels the leg and thinks it's a tree. Each sees only part of the picture.

AI models for diagnosing valvular heart disease (VHD) face a similar challenge. While cardiologists combine information from multiple sources, such as electrocardiograms, imaging data, and patients' reported symptoms, most AI systems still rely on a single data type. As a result, they often lack the holistic perspective needed for clinical decision-making.

In this blog post, relAI PhD student Valentine Idakwo critically examines this issue, identifies the key hurdles, and advocates for changes to enable their practical use.

👉Read the full post: https://zuseschoolrelai.de/blog/your-cardiologist-is-multimodal-why-isnt-your-model/.

We are excited to announce that Shafi Goldwasser, one of the co-founders of modern cryptography, will share her insights at the upcoming Munich AI Signature Lecture. Don't miss the chance to attend her talk!

Title: A Cryptographic Perspective on Machine Learning Trustworthiness Challenges

📅 Tuesday, October 20, 2026, 6:00 pm

📍 Plenary Hall, Bavarian Academy of Sciences and Humanities (BAdW), Munich

About the Speaker

For her groundbreaking work, Shafi Goldwasser received the ACM Turing Award in 2012, the highest honor in computer science; she has also been awarded two Gödel Prizes and is a member of the National Academy of Sciences and the Royal Society.

About the Lecture

In her lecture, Goldwasser explores how cryptographic methods can contribute to trustworthy machine learning.

For decades, cryptographic tools and models have helped make technology platforms trustworthy, even under worst-case assumptions about potential attackers. Goldwasser demonstrates how methods from cryptography and interactive proofs can be used to address trustworthiness challenges at various stages of the machine learning pipeline.

Additional Information

For more info on the lecture and Prof. Goldwasser, please click here

The lecture will be delivered in English

The lecture will also be available via LIVESTREAM on YouTube (no registration needed).

For any question on your participation please contact Annika Mader, events@baiosphere.org

🎉Congratulations!

We are proud to announce that relAI MSc students Nikolai Röhrich and Isabell Hans have received the 🏆 Best Paper Award 🏆 at the German Conference on Pattern Recognition (GCPR) for their paper titled “Concept Guidance: Precise, Training-Free Latent Control for Text-to-Image Generation.”

The article introduces Concept Guidance (CoG), a method that improves control over specific visual concepts in text-to-image diffusion models without requiring retraining, external models, or prompt engineering. 🧐If you are interested in learning more about CoG, you can read the full publication here: https://arxiv.org/abs/2608.14172

👉 More Information about the prize: https://lnkd.in/p/gu6n_yig

On September 22, researchers from relAI, MDSI, and our industry partner SAP came together at the SAP Labs Munich Campus in Garching for the second edition of our Pitchtalks Session.

The event, chaired by Dr. Tobias Müller (SAP), opened with welcoming remarks from Sylvia Kortüm, Managing Director of MDSI, and Dr. Anja Woracek, the new relAI Coordinator. relAI and MDSI researchers then presented their work, followed by lively discussions and networking opportunities.

🌟Highlights from the scientific presentations included:

• Bhavatarini Kumaravel (MDSI) exploring graph-based deep reinforcement learning for structural design

• Annika Schneider (relAI) presenting research on uncertainty quantification in probabilistic models

• Sameer Ambekar (relAI & MDSI): discussing distribution shifts in vision-language models

🤝 The networking session provided an excellent opportunity for researchers and industry experts to exchange ideas, discuss current and emerging challenges in AI, and explore future collaborations.

A big thank you to SAP for hosting and organizing this inspiring event. We look forward to continuing to strengthen the connection between cutting-edge AI research and industry practice in future editions.

🚀 Meet your future employer in AI & Data Science!

We are excited to announce the second Munich Career Fair AI & Data Science 2026 on 29 October 2026 at TranslaTUM, Munich, organized by MDSI, relAI, MCML, and AI HubLMU!

Event Details

📅 Date and Time: 29 October 2026, 2 – 5 pm

📍 Location: TranslaTUM at Klinikum rechts der Isar, Einsteinstraße 25 (Bau 522), 81675 Munich

👉 Agenda and registration: https://go.tum.de/801990

This unique career fair brings together 13 leading industry partners and students from bachelor's, master's, and doctoral programs who are passionate about Artificial Intelligence, Machine Learning, and Data Science.

Take the chance to

🔹 Learn about cutting-edge AI and data science projects
🔹 Discover internships, theses, and career opportunities
🔹 Connect with company representatives directly
🔹 Explore career paths across consulting, technology, research, and industry

 

The new relAI blog post by relAI PhD Student Christopher Bülte explores this important question.

👉You can read it here: https://zuseschoolrelai.de/blog/an-introduction-to-proper-scoring-rules/.

📷 Here’s a snapshot of what you can learn from his post:

Here is a Snapshot of what you can learn about in his post

When predictions come as full probability distributions rather than single values, evaluating them becomes a surprisingly deep problem. This is where proper scoring rules come in.

Proper scoring rules reward forecasts that honestly reflect uncertainty and discourage overconfidence or strategic misreporting. Well-known examples include the Logarithmic Score, Quadratic Score, and CRPS, each providing a different way to measure predictive performance and uncertainty.

What's particularly fascinating is that these scoring rules connect two worlds:

✅ Statistics: Generalizing maximum likelihood estimation and enabling robust parameter inference under model misspecification.

✅ Machine Learning: Serving as training objectives for probabilistic neural networks, improving calibration, robustness, and uncertainty quantification.

As AI systems increasingly move beyond point predictions to probabilistic forecasts, proper scoring rules are becoming a fundamental tool for both model evaluation and learning.

🎉Congratulations!

We are very excited to announce that relAI Fellow Vincent Fortuin has been awarded a European Research Council (ERC) Starting Grant for his project "AutoBayes",  with the title "Unlocking Reliable Small-data AI through Bayesian Deep Learning".

Most current deep learning models require large amounts of training data and tend to be overconfident in their predictions, undermining their reliability. Over the next five years, Vincent and his team will develop Bayesian deep learning methods to create AI models that are confident using limited data while also improving the quantification of the uncertainty in their predictions.

The ERC Starting Grant is one of the most competitive sources of funding available to early-career researchers; it enables them to pursue ambitious projects, to form their own teams and to gain independence at a crucial point in their careers. This year, the funding—amounting to €705 million in total—has been awarded to 421 early-career researchers throughout Europe.

For more information, see the most recent ERC press release.