中文 English Français Русский Español العربية

Why we no longer evaluate SWE-bench Verified

OpenAI Blog2026-02-23 19:000 views

SWE-bench Verified is increasingly contaminated and mismeasures frontier coding progress. Our analysis shows flawed tests and training leakage. We recommend SWE-bench Pro.

Source：https://openai.com/index/why-we-no-longer-evaluate-swe-bench-verified

My favorites

PreviousArvind KC appointed Chief People Officer NextOpenAI announces Frontier Alliance Partners

Qodo builds context into AI code review with Codex

Qodo brings company knowledge and coding standards into Codex, helping developers catch P0 issues across repositories be

0OpenAI Blog

Ford promises not to sell your driving habits to data brokers

Ford vowed to never sell customer driving data to third-party data brokers as part of a new privacy promise aimed at set

0The Verge

Snap settles lawsuit accusing platform of fueling teen depression and eating disorders

The plaintiffs said they started using YouTube as young as age 3 and other social media platforms at around age 9 and de

0TechCrunch

Scrape LinkedIn job postings without logging in (Python + Google Sheets)

LinkedIn shows job postings to anyone, even without an account. In this tutorial we'll collect job postings — with full

0dev.to

Same model, same defences: 7 document attacks down to 0, chat attacks barely moved

A prompt-injection benchmark, three bugs in my own scorer, and the result I only found after fixing them. I built RAGPro

0dev.to

Skill to Create Product Videos

Article URL: https://github.com/aaitor/skill-video-studio Comments URL: https://news.ycombinator.com/item?id=50340992 Po

0Hacker News