Separating signal from noise in coding evaluations
来源原始要点
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.
查看完整报道
直达原文报道 ↗
完整图文、背景资料与讨论归原始发布平台 OpenAI News 所有。