skills.rest
autogluon
autogluonOFFICIAL
Explore all skills by this author →
GitHub

update-leaderboard

Regenerates TabArena website artifacts and refreshes a HuggingFace leaderboard Space's data directory.

313|74|Updated May 15, 2023
Search Tags:#publishing#benchmark#huggingface#leaderboard#artifact generation#git lfs#tabarena
One-click install
npx skills add https://github.com/autogluon/tabarena --skill update-leaderboard-autogluon
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: update-leaderboard
Source: https://github.com/autogluon/tabarena/tree/main/.claude/skills/update-leaderboard
Command: npx skills add https://github.com/autogluon/tabarena --skill update-leaderboard-autogluon

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires huggingface_hub, and includes scripts (resource) components.

What problem does it solve? Publishing new benchmark results to the TabArena leaderboard requires a multi-step, error-prone workflow: regenerating website artifacts, verifying model type classification, safely swapping the Space repo's data directory, and bumping version history. This Skill operationalizes that entire maintainer workflow so nothing is missed. ## Core Features & Use Cases - Model-type pre-check: Verifies newly added models classify into the correct family (Foundation Model, Tree-based, etc.) in website_format.py before regeneration, preventing models from shipping as "Other". - Artifact regeneration: Runs run_generate_website_artifacts.py in the background with monitoring, distinguishing harmless ray shutdown noise from real failures and validating output structure (480 leaderboard CSVs, 8 entrant pools, 2400 explorer HTML files). - Safe data refresh: Deletes the old data/ subtree in the Space repo before copying new artifacts, then verifies the git diff contains only modifications. - Version history bump: Adds a dated entry to website_texts.py with correct verified/unverified model wording. - LFS storage recovery: Includes a HEAD-aware purge script to resolve HuggingFace Space 1 GB storage-limit rejections without breaking the live revision. - Use Case: After uploading a new model's benchmark results, a maintainer asks to refresh leaderboard-testing; the Skill pre-checks the model's type classification, regenerates artifacts, swaps the Space data, bumps the version, and hands off for commit and push. ## Quick Start Update the leaderboard-testing Space with the latest results for the newly added TabSwift model.

Frequently Asked Questions about update-leaderboard

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I update the TabArena leaderboard with new model results?▼

First upload the method's results with the upload-method flow, then run the artifact generator from tabarena/scripts with a tabarena[benchmark] venv, delete the old data subtree in the Space repo, copy the fresh website_data in, and bump the version history in website_texts.py before committing and pushing.

Why does a new model show as Other on the leaderboard?▼

The model's ag_key prefix is not registered in the prefixes_mapping of get_model_family in website_format.py. Add the prefix to the intended family list and a rename entry in get_rename_map, then regenerate the artifacts.

How do I fix HuggingFace Space repository storage limit reached errors?▼

Space repos cap git LFS storage at 1 GB. Run the purge_stale_lfs.py script from within the Space checkout with HEAD at the commit you are about to push; it deletes only LFS objects the local HEAD no longer references while keeping history intact.

Can I preview the leaderboard locally before pushing to production?▼

Yes, use the leaderboard-testing Space as a private preview and start main.py with the Space repo's own .venv, which binds to 127.0.0.1:7860. Confirm it is up by checking the port with curl rather than the log, since gradio buffers stdout.

Does this workflow also update the BeyondArena leaderboard?▼

No, BeyondArena is a sibling flow with its own generator script and a separate data_beyondarena folder in the Space repo. This Skill targets the main leaderboard data directory unless BeyondArena is explicitly requested.

Related Skills

Explore More Software Engineering Skills→
salvo-rssalvo-rs
20

salvo-file-handling

Manage file uploads, downloads, and multipart forms in Salvo web applications.

Official
Intermediate
zero-yxzero-yx

huggingface-git-xet-dataset-publisher

Manages Hugging Face dataset Git repositories with Xet tracking, snapshot pushes, and binary rejection recovery.

Community
Advanced
stellaromicsstellaromics

bio-machine-learning-model-validation

Run nested cross-validation with stratified splits for omics classifiers.

Official
Advanced
surus-latsurus-lat
8

push-to-latamboard

Merge local benchmark results into the HuggingFace LatamBoard leaderboard dataset.

Official
Advanced