SWE-Bench Pro Has a Problem. OpenAI Just Exposed It.
OpenAI’s new analysis reveals serious reliability issues in SWE-Bench Pro, one of AI’s most-cited coding benchmarks. Here’s what broke and why it matters.
OpenAI’s new analysis reveals serious reliability issues in SWE-Bench Pro, one of AI’s most-cited coding benchmarks. Here’s what broke and why it matters.
OpenAI stops using SWE-bench Verified for AI coding tests, citing flawed benchmarks and training leakage. The company now recommends SWE-bench Pro instead.