MLM26: Challenge Showcase

The 4th annual Machine Learning Marathon (MLM26) is a 13-week ML/AI hackathon (Sept 9 - Dec 9) where teams of 3-6 tackle real-world Challenges, share tools and insights, and build skills along the way. Weekly sessions bring workshops, team shareouts, and advisors. Participants also get cloud credits (with training) and access to hosted models. We encourage diverse backgrounds and experience levels to join MLM26; this isn't just for computer scientists, though some challenges may require more expertise than others. Researchers and graduate students are especially encouraged to attend.

Come to the Challenge Showcase on Aug 19, 4:00-5:30pm CT (via Zoom) to learn more. Each organizer pitches their challenge and takes questions, so you can decide on your preferences before you apply. Register to receive the Zoom link, and please forward this to anyone who might be interested!

This year's challenges include: 

• Badger Code: Building a coding agent on relatively small open-weight models, scored on Terminal-Bench
• Badger Scribe: Transcribing 19th-century archival manuscripts with vision-language models and OCR/HTR
• Brain Decoding: Predicting perceived imagery from EEG recordings
• Fixing the Poisoned Well: Auditing a tampered neural network for planted behavior and repairing it, using mechanistic interpretability
• Snapshot WI - Oh Deer: Estimating deer populations from trail-camera imagery
• WattBot 2026: AI+Sustainability literature review and Q&A with retrieval augmented generation (RAG)
   

For more details about prerequisites, visit ml-marathon.wisc.edu/challenges. 

Applications open after the showcase on 8/19 to UW-Madison affiliates with a NetID. For details on cost and how we review applications, visit the FAQ.

We're also recruiting additional advisors and volunteers. The showcase is a good way to see what the teams will be tackling before you sign up to volunteer.

Can't make it? We'll post the recording on the MLM26 site within 48 hours of the event.