Amazon Bedrock announces support for reranker models through the Rerank API, enabling developers to improve the relevance of responses in Retrieval-Augmented Generation (RAG) applications. The reranker models rank a set of retrieved documents based on their relevance to user’s query, helping to prioritize the most relevant content to be passed to the foundation models (FM) for response generation. Amazon Bedrock Knowledge Bases offers fully-managed, end-to-end RAG workflows to create custom generative AI applications by incorporating contextual information from various data sources. For Amazon Bedrock Knowledge Base users, enabling the reranker is through a setting available in Retrieve and RetrieveAndGenerate APIs.
Semantic search in RAG systems can improve document retrieval relevance but may struggle with complex or ambiguous queries. For example, a customer service chatbot asked about returning an online purchase might retrieve documents on both return policies and shipping guidelines. Without proper ranking, the generated response could focus on shipping instead of returns, missing the user’s intent. Now, Amazon Bedrock provides access to reranking models which will address this by reordering retrieved documents based on their relevance to the user query. This ensures the most useful information is sent to the foundation model for response generation, optimizing the context window usage and potentially reducing costs.
The Rerank API supports Amazon Rerank 1.0 and Cohere Rerank 3.5 models. These models are available in US West (Oregon), Canada (Central), Europe (Frankfurt) and Asia Pacific (Tokyo).
Please visit the Amazon Bedrock product documentation. For details on pricing, please refer to the pricing page.
Source:: Amazon AWS