Skip to content

Add LUSR leaderboard submission - #130

Merged
rrivera1849 merged 1 commit into
mainfrom
submit-lusr
Aug 6, 2026
Merged

Add LUSR leaderboard submission#130
rrivera1849 merged 1 commit into
mainfrom
submit-lusr

Conversation

@rrivera1849

Copy link
Copy Markdown
Owner

Summary

Submits the LUSR style embedding model (rrivera1849/LUSR) to the STEB leaderboard.

The goal is to get LUSR onto the public leaderboard under both the Operational and Definitional STEB scores. This follows the standard submission flow described in the README "Submitting your model" section.

What changed

  • SUBMISSIONS.yaml — new entry registering LUSR with its hf_id, run command, and contributor.
  • scripts/models_all.txt — added rrivera1849/LUSR so the default --models-file picks it up during score regeneration.
  • scripts/benchmark_clustering/config.py — categorized LUSR alongside the other style/encoder models so it is scored in the correct group.
  • submitted_results/ — the full LUSR results subtree (metrics.json across every evaluated dataset and task).

Validation

python scripts/validate_submission.py passes: 1 entry declared, 1 short_name on disk, all checks green.

Record the LUSR submission and publish its scores. The model's raw
metrics are merged into the canonical results tarball (distributed via
Google Drive), so this PR carries only the permanent, in-repo artifacts:

- SUBMISSIONS.yaml: register the LUSR submission entry.
- scripts/models_all.txt: add rrivera1849/LUSR to the default models file.
- scripts/benchmark_clustering/config.py: categorize LUSR with the style models.
- docs/leaderboard.md: LUSR now appears in the Operational and Definitional tables.

submitted_results/ is treated as a temporary scratchpad: a contributor's
metrics live there only long enough to be validated and merged into the
tarball, then the subtree is cleared.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant