BlotdleExperts

Authoring tasks for the Blotdle benchmark

Pick a problem from your own work. One you can solve, and one you can check. Write it down here1 and it becomes a task that models are measured against.

The form asks in the terms of your own field. What the data is, what the question is, what you would actually run, and how you would know an answer was right. It does not ask for a Dockerfile2, and it never will.

You give us the specification and the files behind it3. Nothing is built at the moment you submit, so nothing can be run at that moment either. It waits for a person4.