Munich Center for Machine Learning

𝗠𝗖𝗠𝗟 𝗕𝗹𝗼𝗴: Today’s AI systems are remarkably good at recognizing what stands out most. Yet understanding how details relate to each other remains difficult. #MCML Members address this challenge in their work.
🔎 Read more in our blog: mcml.ai/news/2026-02...

#Girls’Day2026: On April 23, 2026, the #MCML, together with Deep Tech Collective and CreAITech , invites girls aged 14–16 to an interactive Girls’ Day at the Hochschule für Philosophie München.

💡 Find out more: www.girls-day.de/.oO/Show/mun...

Yesterday’s #MCML-Stammtisch brought the community together — a perfect chance to exchange ideas and enjoy relaxed conversations beyond the labs 🍻

Join our #MCML Online Seminar Series, where MCML Junior Members present new papers, preprints & ongoing research — everyone who is interested is welcome!

📅 Feb 19, 3:30–4:30 PM
🎤 Dennis Frauen on nonparametric LLM evaluation from preference data
🔗 mcml.ai/events/2026-...

𝗠𝗖𝗠𝗟 𝗕𝗹𝗼𝗴: Trying to find a single moment in a 1-hour video can feel impossible. AI faces the same challenge: models that handle short clips often fail on long videos.

#MCML Members introduce a model that finds exact moments efficiently.

🔎 Read more: mcml.ai/news/2026-02...

𝗠𝗖𝗠𝗟 𝗕𝗹𝗼𝗴: How can machines learn the hidden rules that govern how systems change—how objects move, how weather patterns unfold, or how biological signals evolve—without ever being told what those rules are?

Research from #MCML Members shows that this is possible: mcml.ai/news/2026-01...

💡 𝗠𝘂𝗻𝗶𝗰𝗵 𝗔𝗜 𝗟𝗲𝗰𝘁𝘂𝗿𝗲 with Sebastian Pokutta
🗓️ Thursday, February 12, 2026
⏱️ 05:00-6:30 PM
📍 Geschwister-Scholl-Platz 01, Munich
Find out more: baiosphere.org/en/events/20...

🌍 Research Stay at the University of St. Gallen: Andrea Maldonado spent two weeks at the Institute of Computer Science at the Universität St.Gallen, supported by the MCML AI X-Change Program. 🔎 Read more about her experience: mcml.ai/news/2026-01...

💡 𝗠𝘂𝗻𝗶𝗰𝗵 𝗔𝗜 𝗟𝗲𝗰𝘁𝘂𝗿𝗲 with Lenka Zdeborova
🗓️ Thursday, January 29, 2026
⏱️ 03:00-4:00 PM
📍 Theresienstr. 90, Munich

Find out more: baiosphere.org/en/events/20...

📢 𝗢𝘂𝗿 𝗠𝗖𝗠𝗟 𝗔𝗻𝗻𝘂𝗮𝗹 𝗥𝗲𝗽𝗼𝗿𝘁 𝗶𝘀 𝗼𝘂𝘁! 📢 We are pleased to offer an overview of our research and activities from last year. It reflects the work and commitment of the many researchers involved in our community.
Check it out here: mcml.ai/uploads/mcml...

MCML - From Global to Regional Explanations: Understanding Models More Locally JMLR research shows how subgroup-specific explanations reveal hidden patterns that global model explanations often miss.

𝗠𝗖𝗠𝗟 𝗕𝗹𝗼𝗴: Averaging patterns leads to misleading explanations, e.g., a model predicts bike rentals peak at noon, while in reality, commuters ride in the morning and evening — a common issue in machine learning.
🔎 This is addressed in a paper by #MCML researchers: mcml.ai/news/2026-01...

𝗠𝗖𝗠𝗟 𝗕𝗹𝗼𝗴: Images and text are usually aligned using millions of image–caption pairs. But could they still be matched if they were never seen together?

In “It’s a (Blind) Match!”, MCML Members explore this question.
mcml.ai/news/2026-01...

#MCML launched the #Thinkathon, connecting AI research with real needs in the Bavarian education. MCML Members teamed up to create AI-based teaching kits that schools can actually use.

Thank you for everyone involved for your creativity, expertise, and enthusiasm!

𝗠𝗖𝗠𝗟 𝗕𝗹𝗼𝗴: LLMs are everywhere—but they still struggle to admit uncertainty. Ask a tricky question and they may sound confident, even when they are wrong.

MCML Members show how statistical methods can help LLMs stay honest about what they know (and don’t).

📄 Read more: mcml.ai/news/2025-12...

✈️This summer, Abdurahman Maarouf spent three months as a visiting researcher at Princeton University. In our blog, he reflects on his time abroad — his research, the academic exchange on campus, and life beyond work.

🔍Read more: mcml.ai/news/2025-12...

The #MCML held its annual General Assembly last week, bringing members together for updates, networking, and discussion. A major highlight were the 30+ Flash Talks from Junior Members showcasing their ongoing research.

Great to see the community come together!

#MCML researchers have contributed more than 40 papers to NeurIPS 2025! Check them out: mcml.ai/news/2025-11...

Congrats to our researchers! 🎉 #NeurIPS2025

𝗠𝗖𝗠𝗟 𝗕𝗹𝗼𝗴: Models like CLIP link images and text (“a dog playing with a ball” → 🐕 ⚽ 🌱) and enable zero-shot recognition. But they still struggle with fine details.

MCML members tackle this challenge with more fine-grained, language-guided image representations.

📄 Read more: mcml.ai/news/2025-11...

MCML - Call for Applications: Data Science Professional Certificate Program @LMU 2026 The Data Science Certificate Program@LMU 2026 is a graded extra-occupational training program designed for working professionals and researchers.

Data is everywhere — and knowing how to make sense of it is essential. The Data Science Certificate Program at #LMU in collaboration with #MCML is a part-time academic training course. It is designed for working professionals in academia and industry.
🔎 Find out more: mcml.ai/news/2025-10...

✈️ As part of the #MCML AI X-Change, Kun Yuan visited Stanford and joined Serena Yeung-Levy’s lab in Biomedical Data Science. In our blog, he shares his key learnings, projects, and insights from working at the intersection of AI and biomedicine.

🔎 Read more: mcml.ai/news/2025-11...

𝗠𝗖𝗠𝗟 𝗕𝗹𝗼𝗴: Diffusion models like Stable Diffusion are powerful, but scaling them to large images or videos is still slow and memory-intensive. ZigMa speeds up processing of bigger images with less memory—while preserving visual quality and model expressiveness.

📄 Read more: mcml.ai/news/2025-11...

MCML - Explaining AI Decisions: Shapley Values Enable Smart Exosuits AI meets wearable robotics: MCML and Harvard researchers make exosuits smarter and safer with explainable optimization, presented at ECML-PKDD 2025.

𝗠𝗖𝗠𝗟 𝗕𝗹𝗼𝗴: Businesses are increasingly turning to exosuits to support warehouse workers with repetitive movements. In a new paper, MCML researchers show how these exosuits can be fine tuned with AI.
📄 Blog: mcml.ai/news/2025-11...
🗞️ Börsen-Zeitung: www.boersen-zeitung.de/meinung-anal...

#EMNLP2025 is currently happening in Suzhou, China and #MCML researchers have contributed a total of 37 papers to the conference! 🎉 Check them out: mcml.ai/news/2025-11...

Yesterday’s #MCML-𝗦𝘁𝗮𝗺𝗺𝘁𝗶𝘀𝗰𝗵 was a great opportunity to connect our members across disciplines and enjoy some relaxed conversations outside the labs 🍻

#MCML researchers have contributed a total of 20 papers to #ICCV2025! Check them out: mcml.ai/news/2025-10...

𝗠𝗖𝗠𝗟 𝗕𝗹𝗼𝗴: Deep learning is increasingly helping doctors. However, it’s often hard to understand how these models make decisions.
To change that, #MCML members built SIC, a transparent AI classifier that makes image diagnosis both intuitive and reliable.

📄 Read more: mcml.ai/news/2025-10...

𝗖𝗮𝗹𝗹 𝗳𝗼𝗿 𝗳𝘂𝗹𝗹𝘆 𝗳𝘂𝗻𝗱𝗲𝗱 𝗣𝗵𝗗 𝗣𝗼𝘀𝗶𝘁𝗶𝗼𝗻𝘀: We are offering several PhD positions across our various research areas, open to highly qualified candidates.
‼️ The application portal will be open from 15 October to 14 November 2025.

Find out more: mcml.ai/opportunitie...

Unai Fischer Abaigar is a PhD student at #LMU and #MCML whose work focuses on improving decision-making in public institutions by developing AI systems that are both fair and effective in practice. Check out the interview
👉https://mcml.ai/news/2025-10-09-fischer-abaigar-unai-interview/

YouTube video by MCML_Munich Center for Machine Learning Machine Learning for Climate Action - with researcher Kerstin Forster

Meet Kerstin Forster, researcher at #LMU & #MCML, using machine learning to cut emissions, boost renewable energy, and track corporate sustainability — driving strategies for a healthier planet.

youtu.be/9mYRnXsyvLg

Matthias Feurer, Thomas Bayes Fellow at #MCML & interim prof at #LMU, builds tools to make ML easier for scientists & more efficient for experts. Next semester he starts as junior prof at TU Dortmund—sharing in our interview how the fellowship helped him get there: mcml.ai/news/2025-09...