Rules

The rules of this competition aim to ensure fair evaluation and reproducible results. To this end, we require the following:

• Submissions must be fully reproducible. The top 3 teams with the highest scoring models in each track category are required to submit all necessary code and data to reproduce their model and must be open-sourced and made public after the competition.

• Only the C4 dataset is allowed for both compression and training from scratch. The use of data or content that breaks service contracts or trade secrets of any entity is not allowed. Submissions must NOT use any copyrighted or proprietary data, code, or closed-source content.

• Each team can make three submissions to each track during the competition. The best-performing model will be used to rank the team on the leaderboard and selected for final model evaluation.

• Organizers and advisors associated with the competition are not allowed to participate.

• It is not allowed to mix the submission across tracks. Submitting compressed models to the training-from-the-scratch track will be disqualified, and vice versa.

• This competition will be run under the honor system. Teams that submit very similar results or copy another team’s solution will be disqualified. Violating the spirit of the honor system or taking unfair advantage of the community, even when not against an explicit rule, may result in disqualification and ineligibility for prizes.

• Competition participants can communicate with the organizers to ask questions via various ways such as emails, GitHub issues, and discord channels. Any updates to rules or deadlines will be timely demonstrated on our website and communicated through our discord channels.