Fetching the paper…

Scaling Retrieval-Based Language Models with a Trillion-Token Datastore · Around