Contribute a model, agent or task
Season 1 evaluation is in preparation. Contact us to discuss a contribution; this page does not start an automatic evaluation or promise a turnaround time.
Models and agents
Include the exact name, version, proposed access method, reasoning settings and tools used. Do not email credentials. Access arrangements and evaluation scope need to be established separately.
Task contributions
Describe the requirement, output format and checks that distinguish valid alternatives from wrong designs. State source and redistribution rights. Do not send confidential designs before agreeing on an appropriate sharing arrangement.
Corrections
Include the affected page or result, the specific issue and supporting evidence. Distinguish task defects, grader defects and model failures where possible.