Submit-answer is underspecified

[Matistjati]

Question 1: do we want to allow code submissions? I don't think it's completely unreasonable. IOI 2025 had one task where you could either submit code or answer files.

Question 2: how do you write sample solutions? If code is allowed, that's half the equation. But should we perhaps allow submissions in the form of a directory filled with .ans files?

I think it's very important to clarify these. Especially, if code is not allowed, then we need a way to test submissions. This is highly relevant for the Swedish AI olympiad.

[mzuenni]

currently the spec says:

A submission consists of the answers to the test cases instead of source code for a program that produces the answers.

so no submissions. The example on the other hand here only has code as submissions...

I have another question: Can there be multiple test cases? If yes, how is a multi file submission supposed to work? (how are submitted files mapped to testcases)

[Matistjati]

After discussing with @niemela (a while ago), I think we want the following:

• There can be multiple test cases, along with output validators
• Submissions can come in many forms, and is probably outside the scope of the spec, except for sparking the imagination of the implementer.
  • Via a Kattis-like interace (see right box here https://open.kattis.com/problems/lvable) where you type your answer and submit
  • Upload a zip containing file names that match test case names
  • Drag and drop a subset of .ans files
  • Uploading a submission that is run the .in files (recall that these must exist to define a test case)
• Judge solutions can either be in the form of code, or a folder with a set of .ans files
• I don't think we want to allow having no judge solution, even if there exists a trivial one (the .ans files in secret)
• Did I forget anything else

How would you prefer to structure this? We can either create a new submit-answer section, or scatter the information where relevant. It seems that we prefer scattering the info?