Tests whether your benchmark leaked

Your benchmark went up two and a half points because the model had seen the answers

You cannot inspect the training data, and you do not need to. If memorised items score better than fresh ones by more than chance, the number you are reporting is inflated — and that is a test, not an opinion.

No spam. One email when it is ready to try.

The problem

Eval credibility rests entirely on trust, and the trust is starting to go

Every model release quotes benchmark numbers nobody can independently check, and every serious practitioner privately assumes some contamination. There is no accepted way to measure it, so the argument stays at the level of suspicion — which is bad for the labs being doubted and worse for the ones being honest.

22 of 180
items in our benchmark lifted from a public repository
2.5pp
inflation in the headline number they cause
zero
access to training data required to detect it

The insight

Contamination is a two-group comparison, not a search of the corpus

Split items by how much they overlap public text, compare scores between the groups, and ask whether the gap could have arisen by chance. A permutation test answers that with no distributional assumptions and no access to anything private. If the split is uninformative the p-value says so, which makes a clean result meaningful rather than merely unfalsified.

Method

N-gram overlap scoring against public corpora to partition items, mean-score gap between suspect and clean groups, and a permutation null over label assignments for an exact p-value, with a contamination-adjusted score reported alongside the raw one.

How it works

Four steps, no data science team

01
Score the overlap

Every eval item gets an overlap score against public text.

02
Split and compare

Suspect against clean, on the model you are evaluating.

03
Permute

Thousands of relabellings give the null, and the p-value follows.

04
Report both numbers

What you measured, and what it would be on clean items only.

Who it is for

Whoever puts their name on the eval result

Labs, eval vendors and anyone publishing a benchmark number they expect to be challenged on.

Pricing

Free
$0
One benchmark, one model, published methodology.
  • Overlap scoring
  • Permutation test
  • Adjusted score
Most common
Team
$900/mo
Continuous checking across benchmarks and model versions.
  • Continuous checking
  • Version comparison
  • Corpus updates
  • Certificate export
Lab
Custom
Self-hosted, with private corpora and custom overlap measures.
  • Self-hosted
  • Private corpora
  • Custom overlap
  • Direct support

Competition

What exists, and what it does not do

WhoWhat they doThe gap
Canary strings and held-out setsBenchmark authors mark or withhold data to detect leakage.Only works if you own the benchmark and planned ahead. Useless for anything already published.
Contamination researchAn active academic literature with several detection methods.Papers rather than a service, and no standard a third party can point at when challenged.
Private held-out benchmarksKeep the test set secret.The strongest defence and it does not scale — the secret set leaks the moment it is used enough to matter.
Assuming it is fineCurrent practice.Increasingly untenable, and the first serious public dispute will end it.
How this fails

The buyer set is small and unusually sophisticated — labs employ people who have read every contamination paper and can implement this themselves, so the sale is about neutrality rather than capability. That points at being a third-party certifier rather than a vendor, which is a slow, credibility-first business with almost no revenue in year one. And overlap is a proxy: a model can memorise something without lexical overlap, so a clean result is weaker evidence than a dirty one, which must be said plainly or the product is dishonest.

Market

Small in dollars, disproportionate in credibility — it is the reference everyone else cites

A few dozen serious buyers. Better read as an authority-building asset that makes Eval Kits and the Index credible than as a business on its own.

Test my benchmark

No spam. One email when it is ready to try.

Or just go look at the demo first →