zerank-2-reranker

Name: zerank-2-reranker
Author: zeroentropy

by zeroentropy

SOTA reranker trained with Elo-based distillation from relevance judgments

164Kdl/month

4Bparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

zeroentropy/zerank-2-reranker

Feature URI

mixpeek://reranker@v1/zeroentropy_zerank2_v1

Overview

ZeRank-2 from ZeroEntropy is a 4B parameter reranking model that achieves state-of-the-art accuracy across diverse retrieval benchmarks. It outperforms Cohere Rerank v3.5 and Gemini 2.5 Flash on standard retrieval tasks with an average nDCG@10 of 0.6714 on MTEB v2. The model uses a novel Elo-based training approach where relevance is calibrated through pairwise preference distillation, producing more accurate relevance scores than traditional pointwise cross-encoders.

Architecture

Cross-encoder built on Qwen3-4B, fine-tuned with Elo-based distillation. Processes concatenated query-document pairs through the full transformer stack and produces a scalar relevance score. The Elo training approach uses a teacher reranker to generate pairwise preferences, then trains the model to predict calibrated Elo ratings as relevance scores. Supports 32K token context.

Mixpeek SDK Integration

// Reranking is a retriever STAGE in Mixpeek, not an ingest-time extractor.
// The rerank stage runs a cross-encoder inference service; the shipped default
// is BAAI/bge-reranker-v2-m3. Pointing it at zerank-2-reranker means registering that
// model as a custom reranker plugin and naming it in feature_uri, which is an
// Enterprise path. Stage contract read from GET /v1/discovery/stages.
const retriever = await mx.retrievers.create({
  namespace_id: "my-namespace",
  retriever_name: "search-then-rerank",
  stages: [
    {
      stage_name: "candidates",
      stage_id: "feature_search",
      parameters: { limit: 100 },
    },
    {
      stage_name: "rerank_results",
      stage_id: "rerank",
      parameters: {
        inference_name: "BAAI__bge_reranker_v2_m3",
        query: "{{INPUT.query}}",
        document_field: "content",
        top_k: 10,
      },
    },
  ],
});