Living benchmarks and leaderboards for machine learning on tabular data
"""
INTRODUCTION_TEXT = """
**TabArena** is a living benchmark for predictive machine learning on IID tabular data, built to measure the
peak performance of model-specific pipelines.
"""
# Conflict-of-interest statement, surfaced as a small corner hint that opens a
# CSS-only popup (see main.py). Condenses our public position: name the conflict
# plainly, then make the case β open code, open evidence, competing institutions β
# that it does not affect the validity of the results. No JS so it works inside
# the embedded Hugging Face Space iframe.
COI_HTML = """
π Our Conflict of Interest
TL;DR: TabArena maintainers also develop models and are affiliated with
commercial institutions. This creates a conflict of interest. We do not hide it, and we believe (and let
anyone verify) that it does not affect the validity or scientific rigor of the results shown here.
Where the conflict lies
Maintainers vs. model developers. Several of us develop tabular models that appear on this
leaderboard. Building the benchmark and competing on it at the same time is a tension we can't fully
remove (it is also documented in the Competing Interests section of our NeurIPS paper).
Science vs. industry. Some maintainers are employed by commercial institutions. We are also
open-source researchers and PhD students maintaining TabArena in the open.
We won't pretend that simply telling you we act without bias settles the question; it doesn't, and you
are right to stay skeptical. And we'll say it outright: maintainers building a benchmark that features
their own models is not, by itself, a prudent arrangement.
Why we believe the results still stand
Rather than ask you to trust us, we try to make trust unnecessary by letting anyone check everything:
Everything is open source: the benchmarking code, the leaderboard, the plots, the model
implementations, and the models themselves.
So is the evidence: we publish the raw predictions, hardware details, and exact software
environments. Anyone can reproduce our numbers, hunt for errors, submit their own model, and call us
out.
This puts the burden of proof on the other side: we hand any critic everything they would need to prove
we did something wrong. Almost no one will check all of it, but the point is that anyone can.
Competing interests as checks and balances
TabArena is maintained, on purpose, by people from different and competing institutions, whose
interests pull against each other. That way no single model or company sets the rules unchallenged, and we
keep widening that mix as the benchmark grows.
What we are still working on
TabArena is young and still changing fast. The biggest thing we're missing is governance: we'd like
proper open-source governance, and we have a TabArena organization and are heading that way, but right now
we're too few people with too little time to set it up. We bring in more of the community with every
release, and we'll keep adjusting.
"""
# Jump menu, built like COI_HTML: a checkbox toggles a CSS-only popup, so it needs no JS and
# works inside the embedded Hugging Face Space iframe. Targets are the section anchors set in
# pages.render_internal_page.
TOC_HTML = """