正在加载视频...

视频加载失败

Help shape the future of Android Bench. Android Bench provides you with transparency into AI coding assistance. You now have an opportunity to contribute your own tasks, or submit locally run evaluations. Read more to learn how →

19,483 次观看 • 14 天前 •via X (Twitter)

2 条评论

MD ADIL 的头像
MD ADIL14 天前

Update it with latest model, 3.8 Flash specifically.

PuneetSingh⚡️ 的头像
PuneetSingh⚡️14 天前

locally run evals is the useful bit. once people submit their own tasks, how do you stop the set from overfitting — versioned harness + held-out tasks?

相关视频