Brett Winton: 是否有人尝试通过语言复杂性（或以其他方式对数据进行逻辑排序）对数据进行批量排名，以此来提高训练效率？

原推：Has anybody tried batch ranking data by language complexity (or otherwise logically sequencing data) as a way to improve training efficiency?
https://twitter.com/wintonARK/status/1623908468452384768