Moreover, you get surprisingly out-of-class (size-wise) performance if you fine-tune for your specific problem space. Even if you only train in a parameter-effi

Moreover, you get surprisingly out-of-class (size-wise) performance if you fine-tune for your specific problem space. Even if you only train in a parameter-efficient way.