Claude Skills by autogluon

Adapt TabArena/bencheval as the tasks/models/metrics/splitting layer for a new, domain-specific benchmark that lives in its own repository (not a contribution to this repo). Use this skill whenever the user wants to build a benchmark for a different data domain (e.g. spectroscopy, genomics, time series, a vertical-specific tabular task) on top of TabArena's model zoo, task and experiment runner, and bencheval's leaderboard math, rather than reimplementing that layer from scratch. Triggers on ...

Add a new ML model to the TabArena benchmark system. Use this skill whenever the user wants to integrate a new tabular ML model into TabArena — even if they just say "add X model", "integrate X", "support X", or "wrap X for the benchmark". Creates all required files: the AutoGluon model wrapper, the search-space generator, the per-model `info.py`, and the `pyproject.toml` extra (the model is fit-tested automatically by the registry-driven `test_all_models.py` — no per-model test file). Reads ...

Add a new ML *system* to the TabArena benchmark. Use this skill whenever the user wants to integrate a whole pipeline rather than a single model — AutoML frameworks (AutoGluon, LightAutoML, FLAML, auto-sklearn), LLM-driven agents, hosted prediction APIs (TabPFN-3-API), or a model run through a heavier self-managing interface (TabFM+). Triggers on "add X as a system", "benchmark the X AutoML framework", "wrap this API for TabArena", "integrate this agent". Creates the per-system folder (`syste...

Run one already-integrated model on the TabArena benchmark, from a local smoke fit to the cluster run and the evaluated leaderboard. Use this skill whenever a maintainer wants to benchmark a model that is already in the registry, e.g. "benchmark TabM", "run Nori on the cluster", "create a setup/eval script for DenseLight", "launch <model> on TabArena and evaluate it". It first asks whether Claude should drive the run end-to-end here (launch, monitor, evaluate, report) or hand the launch and m...

Regenerate the TabArena website artifacts and refresh a leaderboard Space's `data/` with them. Use this skill whenever a maintainer wants to publish the latest results to the leaderboard — e.g. "update the leaderboard", "regenerate the website artifacts and refresh the LB", "push the new results to the leaderboard Space", "refresh leaderboard-testing with the latest". Runs `scripts/run_generate_website_artifacts.py`, pre-checks that any newly added *models* classify correctly (Foundation Mode...

Process, upload, and register one benchmarked method's results in TabArena. Use this skill whenever a maintainer points at a benchmark run's output directory and wants to turn its raw `results.pkl` files into cached + hosted + registered TabArena artifacts — e.g. "upload this method", "process and upload these results", "host/publish <model>'s results", "register <model> in the leaderboard". By default Claude runs the whole flow itself (inspect → fix `info.py` → process → upload dry-run → rea...