dailyai.report

The Ettin reranker family optimizes retrieval-augmented generation by refining document ranking. These models use a cross-encoder architecture to score the relevance of retrieved passages more accurately than standard bi-encoders. This reduces noise in RAG pipelines.