评测与研究

jev-korean-benchmark

@mahlernim5Python更新于 2026-09-17

对 Jev 在韩语理解和医学文本上的早期评测,包含运行时和成本数据,支持复现。

英文原文

Reproducible early-access evaluation of Jev on Korean understanding and medical text, with runtime and cost evidence

mahlernim/jev-korean-benchmark

它在哪儿调用了 Jev

from typesafe_sdk import Choice, RetryPolicy, TypeSafeClient

jevbench/medqa_run.py:10

链接指向我们抓取当天的那个 commit,行号是准的。

这个项目没法在这儿跑

它的 question 组合是运行时拼出来的,或者代码里没有直接写出来,所以没法原样搬过来。源码链接在上面,可以自己去看。

同类的其他项目