← Back to feed

How I made a benchmark mined from my own repo that agreed with FrontierCode

AI · 2026-10-21 14:10

作者从自己仓库挖掘24个任务构建基准，模型排名与FrontierCode相关系数达0.96，认为私有化基准更能反映真实使用体验。

Source: www.stet.sh

Auto-generated summary for reference only. RSS