Por que não avaliamos mais o SWE-bench Verified

O SWE-bench Verified está cada vez mais contaminado e mede incorretamente o progresso da codificação de fronteira. Nossa análise mostra testes falhos e vazamento no treinamento. Recomendamos o SWE-bench Pro.

Fonte: OpenAI Blog

Publicado em 2026-02-23

Notícias relacionadas

• From Enterprise to Earthshot: UK’s Ecosystem Moves From Ambition to Application With NVID… — Across industries, the U.K. is putting the full stack to work — and the scale of deployment is growing fast, from enter…
• European Cities Deploy NVIDIA-Powered AI Agents to Run Smarter Urban Operations — Keeping a city running means making decisions that affect residents every day — from when to repair a road to how to us…
• Google Releases Gemini 4 Argon API With 1M Token Output Limit — Google has made Gemini 4 Argon generally available on the Gemini API, AI Studio and the Gemini app. The sparse MoE mode…
• Meta Launches Muse Wonder and Muse Wonder Full: Benchmarks, Pricing and How to Use It — Meta released Muse Wonder and Muse Wonder Full, 1M-context agent models. Muse Wonder costs $4/$20 per million tokens on…
• How to Use Meta’s Rebalancer for Task Placement & Optimization — In this comprehensive coding tutorial, we explore Meta's Rebalancer assignment solver using its C++ core and Python bin…
• All Systems Go: Global PC App Ecosystem Ready for NVIDIA RTX Spark — NVIDIA RTX Spark is a new beginning for the PC — a reinvention combining the capabilities of a gaming PC, a creative st…

Continue explorando

• Ver todas as notícias de IA
• Catálogo de ferramentas de IA
• Cursos e tutoriais de IA