Contribute

Propose a new model or dataset. Repository and submission links are withheld during double-blind review.

How it works

From proposal to leaderboard

  1. 1
    On this page

    Propose

    Fill in the form below.

  2. 2
    GitHub issue

    Discuss

    Maintainers confirm scope and ID.

  3. 3
    Pull request

    Implement

    Anyone opens a pull request.

  4. 4
    After merge

    Benchmark

    Results appear on the leaderboard.

Already have code? Skip to the implementation guide.

Step 1 · Propose

Describe your proposal

AWhat would you like to add?

BWho will implement it?

CTell us about it

Step 3 · Implement

Implementation guide

Code and tests only. No results, no raw data.

  1. 1

    Add the implementation

    Fork [Model repository] and add src/whar_models/<model_id>/ with a builder taking input_channels, window_length and num_classes.

  2. 2

    Register it

    Add it to WHARModelID and ModelSpec in registry.py.

  3. 3

    Test it

    Add a forward-pass test for varying channel, window and class counts.

    uv sync --extra dev uv run pytest
  4. 4

    Open the pull request

    Link the proposal with Closes #<issue>.

    [Pull request link]
Step 4 · After merge Maintainers benchmark the entry and publish the results.