用 Jev 分类模型逐步选择合法 token 编写 C++
作者用 TypeSafe 的 Jev 决策/分类模型做实验:每步只提供语法合法的候选 token,让模型选择,从而生成 C++ 代码。
Ronald te Brake 分享使用 Jev 进行语义测试排序的实验:在 +1,394/-39 的差异中,评估 852 个测试文件,仅选择 17 个运行,Jev 输入成本约 $0.41,减少 98% 的测试。
Experimenting with Jev for semantic test ranking in our test impact analysis. One diff of +1,394 / -39: 852 test files assessed 17 selected to run ~$0.41 Jev input cost That's 98% fewer proposed tests. I manually checked the selection, and these were the tests I would have