Biological generative models can predict biological functions without task-specific training data but often under-perform specialized models. This is due to a fundamental 'alignment gap', where the rules learned during unsupervised training are not related t... ...

请注册登录后继续浏览