Jev 在分类器评测中质量与速度均胜出
Malte Ubl 将 TypeSafe AI 的 Jev 用于现有分类器评测,结果在质量上饱和评测集,速度比 Gemini 2.5 Flash Lite 快 6 倍。
Nathan LeClaire 提出,Jev 能快速并发执行大量评估,因此可以拆分、变异并实验输入,存储数值结果,从而实现显著性分析。
Here's food for thought @simonw one thing unique about jev is that you can do so many evaluations so fast (incl. concurrently) that you could break apart, mutate and experiment with inputs and store the numeric results, this probably would allow some 'significance analysis'