Source Prompt: Coordinated Pre-training of Language Models on Diverse Corpora from Multiple Sources

Open in new window