rerank

What problem does it solve?

This skill improves the relevance of search results by reranking initial results with a cross-encoder.

Core Features & Use Cases

• Second-stage ranking: Use cross-encoder reordering to refine top-K results after an initial vector search.
• Flexible integration: Works with existing pipelines (RAG, semantic search) to boost precision without rebuilding recall.
• Use Case: When initial results are broad, apply rerank to surface the most relevant documents for a query like "What is Milvus?".

Quick Start

Use the rerank skill to re-order the top-K results from a vector search by applying a cross-encoder re-ranker model to the candidate docs.