Borrowing from the law to filter training data for foundation models

Using a “Pile of Law" dataset, Stanford researchers explore filtering private or toxic content from training data for foundation models.Read More