OpenAI News·· 2026-07-08AI 评分35
OpenAI 分析揭示 SWE-Bench Pro 编码基准问题
Separating signal from noise in coding evaluations
AI 导读
OpenAI 新分析指出，流行的编码基准 SWE-Bench Pro 存在问题，引发对评估 AI 模型可靠性和准确性的担忧。
原文
A new analysis from OpenAI reveals issues in SWE-Bench Pro, a popular coding benchmark, raising concerns about reliability and accuracy in evaluating AI models.
来源：OpenAI News · openai.com