Don't trust the model, trust the process
A look at how Meedan evaluates Suwali before it reaches users, from golden datasets to automated retrieval and generation metrics and blind review by the native-speaker authors who wrote the questions.
A look at how Meedan evaluates Suwali before it reaches users, from golden datasets to automated retrieval and generation metrics and blind review by the native-speaker authors who wrote the questions.
Funded by:
