A Fresh Take on Stale Embeddings: Improving Dense Retriever Training with Corrector Networks

Open in new window