Rerankers Overview

Improve search result quality by reordering results after initial retrieval.

What are Rerankers?

Rerankers are powerful post-processing operations that improve the quality and diversity of search results by reordering them after the initial retrieval phase. They enable you to:

• Combine results from multiple search strategies (hybrid search)
• Reduce redundancy by diversifying results
• Optimize the relevance-diversity trade-off
• Improve the overall user experience of your search application

When to Use Rerankers

Apply rerankers in your query pipeline when you need to:

• Merge multiple search methods: Combine vector search with BM25 keyword search, or merge results from multiple vector searches
• Diversify results: Eliminate near-duplicate content and show varied perspectives
• Optimize ranking: Fine-tune the balance between relevance and variety based on your use case
• Improve search quality: Leverage sophisticated ranking algorithms without changing your underlying search infrastructure

Available Rerankers

HelixQL provides two powerful reranking strategies:

RerankRRF (Reciprocal Rank Fusion)

A technique for combining multiple ranked lists without requiring score calibration. Perfect for hybrid search scenarios where you want to merge results from different search methods.

::RerankRRF             // Uses default k=60
::RerankRRF(k: 30.0)    // Custom k parameter

RerankMMR (Maximal Marginal Relevance)

A diversification technique that balances relevance with diversity to reduce redundancy. Ideal when you want to show varied results instead of similar or duplicate content.

::RerankMMR(lambda: 0.7)                           

Basic Usage Pattern

RRF Usage:

QUERY SearchDocuments(query_vec: [F64]) =>
    results <- SearchV<Document>(query_vec, 100)
        ::RerankRRF             // Apply reranking
        ::RANGE(0, 10)          // Get top 10 results
    RETURN results

MMR Usage:

QUERY SearchDocuments(query_vec: [F64]) =>
    results <- SearchV<Document>(query_vec, 100)
        ::RerankMMR(lambda: 0.7)  // Apply reranking
        ::RANGE(0, 10)            // Get top 10 results
    RETURN results

Chaining Rerankers

You can chain multiple rerankers together for complex result optimization:

QUERY AdvancedSearch(query_vec: [F64]) =>
    results <- SearchV<Document>(query_vec, 150)
        ::RerankRRF(k: 60)       // First: combine multiple rankings
        ::RerankMMR(lambda: 0.6) // Then: diversify results
        ::RANGE(0, 10)
    RETURN results

Best Practices

1. Retrieve more results initially: Fetch 100-200 candidates to give rerankers sufficient options to work with
2. Apply rerankers before RANGE: Rerank first, then limit the number of results returned
3. Choose the right reranker: Use RRF for combining searches, MMR for diversification
4. Test with your data: Experiment with different parameters to find what works best for your use case

Common Patterns

// Pattern 1: Simple diversification
SearchV<Document>(vec, 100)::RerankMMR(lambda: 0.7)::RANGE(0, 10)

// Pattern 2: Hybrid search fusion
SearchV<Document>(vec, 100)::RerankRRF::RANGE(0, 10)

// Pattern 3: Fusion + diversification
SearchV<Document>(vec, 150)::RerankRRF::RerankMMR(lambda: 0.6)::RANGE(0, 10)