Code of Conduct
To keep this a collaborative and friendly learning environment, all participants and advisors are expected to follow the Code of Conduct below.
- Embrace the Kaggle spirit. Kaggle thrives on open collaboration. Share your work, your notebooks, and what you’ve figured out. Solutions move between teams here all season and that’s the point.
- Give credit. If an approach, a snippet, or an idea came from another team, a notebook, or a paper, say so. Sharing works because attribution is automatic.
- Respect and inclusivity. Encourage a range of perspectives and backgrounds. Avoid offensive language or comments that would make others uncomfortable.
- Be patient and supportive. People arrive with very different amounts of experience. Offer help and mentorship when it’s wanted, and explain terms others may not know.
- Feedback and critique. Give constructive feedback respectfully, and be open to receiving it on your own work.
If something goes wrong. If you experience or witness behavior that violates this Code of Conduct, contact Chris Endemann at endemann@wisc.edu. Reports are handled discreetly. Depending on what happened, the response may range from a conversation to removal from the Marathon without refund.
Rules
In addition to the Code of Conduct, all participants are expected to follow the rules below.
1. Team size. Teams must have 3 to 6 members when they form. This range is enforced at formation only. If a team loses someone mid-season, the team continues as is. Teams should be finalized by the 9/9 kickoff. Mergers may be requested and are typically approved if both teams agree.
2. Mixer attendance. Plan to attend the 9/2 mixer. If you have a conflict, post in your challenge’s Slack channel by 8/31 – or within 24 hours of receiving your ticket, if you’re admitted after that – and work with the channel to find a team asynchronously. The channel name and the Slack invite come with your ticket confirmation. If you neither post nor show up, your seat is released to the waitlist.
3. One challenge per participant. So efforts aren’t spread too thin, each participant competes in one challenge, on one team.
4. In-person participation. Marathon participation is in person. Most sessions are at Orchard View 3280, 330 North Orchard St., Madison; a few are elsewhere, including the 9/16 workshop at Morgridge Hall 7560. See the schedule for locations. The challenges themselves are public on Kaggle and anyone may work them independently, but that is not Marathon participation and does not come with a seat, advisors, compute, or access to in-person sessions.
5. Attendance. Participants are expected to attend nearly all scheduled events. Occasional absences (exams, unavoidable conflicts) are understandable, but be respectful of the commitment your teammates have made. Arriving after 5PM is fine as long as your team approves. Presentations and workshops typically don’t start before 5PM.
6. Deliverables. Every challenge requires a writeup. Deadlines are the same across challenges:
- 9/28: lightning talk slides due
- 11/10: draft writeup due
- 11/17: peer comments on other teams’ drafts due
- 12/2, 11:59PM: final writeup due and public leaderboards close
7. Challenge-specific rules. Review the Rules section of your challenge on Kaggle. Individual challenges may add constraints – open-weight models only, hardware limits, dataset restrictions – and those apply on top of these rules.
8. Coding agents and AI assistance. Allowed and expected; we run a workshop on it. Two conditions: you’re responsible for verifying what a model produces before it goes in your submission, and your writeup should say which tools and models you used and where they did the work.
9. Shared compute. Each participant gets roughly $50 in AWS credits drawn from a shared pool, and some challenges get access to hosted campus LLMs. The pool is finite and shared across all teams, so use it deliberately. Prototype locally before scaling up. Credits are for Marathon work only.
10. Eligibility. Participants must be UW-Madison affiliates with a working NetID for the duration of the Marathon. NetID access is required for cloud compute and hosted models. See the FAQ for how applications are reviewed.
11. Minimum requirements. All participants are expected to know Python, machine learning fundamentals, and enough GitHub to collaborate. If you need to get up to speed, complete the relevant self-paced workshops before kickoff: Intro Python, Collaborating with GitHub Desktop, and/or Intro to Machine Learning.
12. Registration and refunds. Refunds (minus $25 processing fee) are available through 9/3, 11:59PM – one day after the mixer. After 9/3 the ticket is non-refundable. See the FAQ for cost, funding string, and waitlist details.