Isaacus

Reranking

Put the most relevant legal context first

Overview

The second pass your retrieval pipeline needs.

Reranking takes a broad set of candidate results and reorders them by true relevance to a query. Use it after search or embedding retrieval to make legal RAG systems more precise, more useful, and less likely to send weak context into the model.

Why it matters

Better context in. Better answers out.

Kanon 2 Reranker helps legal AI products choose the laws, decisions, contracts, evidence, and passages that are most likely to matter.

  • Put the best context first

    Rerank candidate passages, laws, decisions, contracts, evidence, and internal documents so your RAG system gives the model the material that actually answers the query.

  • Better legal retrieval

    Paired with Kanon 2 Embedder as a first-stage retriever, Kanon 2 Reranker delivers 18% better performance on Legal RAG Bench and 6% better performance on MLEB.

  • Works with long legal documents

    Kanon 2 Reranker has first-class support for documents of any length, powered by our semchunk semantic chunking library.

How teams use it

From rough retrieval to ranked legal evidence

Add reranking to legal search, research copilots, contract review tools, evidence workflows, and RAG systems that need higher-quality context.

  • RAG that uses the right evidence

    Improve the quality of legal AI outputs by making sure the strongest retrieved context reaches the model first.

    Most RAG failures start before generation. The system retrieves something plausible, but not quite the right authority, clause, exhibit, or passage. Reranking gives your retrieval pipeline a second pass before the model starts writing.

    Kanon 2 Reranker scores candidate documents against the user’s query and reorders them by true legal relevance. That means better context for research copilots, contract review tools, evidence analysis, knowledge search, and agentic legal workflows.

    For legal engineers and vibe coders, the pattern is simple: retrieve a broad candidate set, rerank it, then pass only the highest-signal context into the next step of your application.

    • RAG
    • Retrieval
    • Legal engineering

  • Kanon 2 Reranker

    A legal reranking model that reorders retrieved documents by how relevant they really are to a query.

    As the first reranker optimized specifically for legal RAG, Kanon 2 Reranker excels at scoring the relevance of laws, decisions, contracts, evidence, and other legal documents to legal queries.

    When paired with Kanon 2 Embedder as a fast, affordable, and accurate first-stage retriever, Kanon 2 Reranker delivers 18% better legal information retrieval performance on Legal RAG Bench and 6% better performance on the Massive Legal Embedding Benchmark (MLEB).

    It is built for production retrieval systems where the first search pass is not enough. Use it after embedding search, keyword search, hybrid search, or another candidate generator to decide which results deserve to be shown, cited, summarized, or passed into a model.

  • Long-document reranking

    Score long legal documents without forcing builders to pretend every useful answer fits inside a short chunk.

    Legal relevance often depends on more than a single sentence. A contract clause may need its definitions, a case may need its facts and holding, and evidence may need surrounding context before it can be judged properly.

    Kanon 2 Reranker supports documents of any length using semchunk, our semantic chunking library. That lets your application work with long contracts, decisions, filings, bundles, policies, and matter documents while still ranking the pieces that matter most.

    The result is a retrieval layer that is easier to trust: broad enough to catch the right material, precise enough to put the best context first, and practical enough for real legal document collections.