评测与研究

jev-benchmark

@wondertwins2PythonMIT更新于 2026-09-16

TypeSafe 的 Jev 模型基准测试和试玩平台,包含国际象棋和语音转文本游戏 NPC 对话识别。

英文原文

Benchmarks and a playground for TypeSafe's Jev (System One) model: chess, and who-is-the-player-talking-to for speech-to-text game NPCs

wondertwins/jev-benchmark

它在哪儿调用了 Jev

from typesafe_sdk import Choice, Noul, Score

jevchess/experiments.py:22

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