Observed arrival · 2026-08-20
Goldset: Bug Fixes You Can Actually Reproduce
A dataset of real Python software bugs paired with their maintainers’ fixes and tests that fail before the fix and pass afterward.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
Why it surfaced
Goldset keeps a record only after running both sides of the change, and its method uses no language model. It also separates a public corpus from 385 held-out records, giving coding-agent evaluators a concrete answer to the contamination problem.
A dataset of real Python software bugs paired with their maintainers’ fixes and tests that fail before the fix and pass afterward.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
ƒJavaScriptBrowser-side code central
One card from the complete issue