ProductJAF判断框架
JAF
An online framework for agent-based judgment tasks, supporting multilingual intent recognition and other classification workloads. It serves as a comparative baseline for local decision models such as Laya.
Timeline (last 90 days)
Sep 25
Laya achieves only 0.45 accuracy on the MASSIVE dataset covering 13 languages and only 0.425 on Banking77, while JAF reports 0.87 on the same tasks.
Unverified50%