Boosting Your Search and RAG with Voyage’s Rerankers

Rerankers, like Voyage's new rerank-lite-1, are neural networks that improve the quality of search results in Retrieval-Augmented Generation (RAG) systems by scoring and re-ranking initial search outcomes based on relevance. Voyage rerank-lite-1 has been evaluated across 27 datasets, demonstrating superior performance compared to competitors like bge-rerank-large and Cohere’s rerank-english-v2.0 in fields such as technical documentation, law, and medicine. The reranker works as a refinement step in a two-stage retrieval system, where it uses cross-encoder neural networks to analyze the interactions between queries and documents more intricately than embedding-based methods. Although this approach offers enhanced accuracy, it incurs higher computational costs. Voyage rerank-lite-1 supports a 4K-context length, offering flexible pricing based on token usage, and is available as an API endpoint and Amazon Marketplace Model Package. The reranker consistently improves recall metrics when integrated with various first-stage search methods, except in cases involving code data, where other embeddings perform better.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.