ReMatch-3B

Name: ReMatch-3B
Author: FireRedTeam

by FireRedTeam

Multimodal retriever trained with generative matching for stronger query-item alignment

14dl/month

5likes

3Bparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

FireRedTeam/ReMatch-3B

Feature URI

mixpeek://image_extractor@v1/fireredteam_rematch_3b_v1

Overview

ReMatch turns a multimodal LLM into a retrieval model by adding a chat-style generative matching objective. Instead of relying only on contrastive pairs, it teaches the model to reason about whether a query and candidate match, then distills that signal into retrieval embeddings.

On Mixpeek, ReMatch is relevant for agent retrieval when queries are specific, compositional, or visual-textual, such as finding a frame where a person is doing one action while an object appears in a certain place.

Architecture

3B multimodal retriever with learnable representation tokens and a generative matching training objective. The model supports English and Chinese according to the model card.

Mixpeek SDK Integration

// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so ReMatch-3B runs
// on your side and the output is upserted through POST
// /v1/namespaces/{namespace_id}/documents/upsert. On Enterprise the other
// path is to upload the weights instead: POST /v1/namespaces/{id}/models
// accepts the huggingface format and a custom plugin loads them.
const res = await fetch(
  "https://api.mixpeek.com/v1/namespaces/ns_your_namespace/documents/upsert",
  {
    method: "POST",
    headers: {
      Authorization: "Bearer API_KEY",
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      collection_id: "col_your_collection",
      documents: [
        {
          document_id: "asset-00412",
          // The vector name has to match a vector index on the collection.
          vectors: { "multimodal-embedding": yourVector },
          payload: { source_key: "archive/2026/asset-00412" },
        },
      ],
    }),
  },
);

// Managed alternative, if this exact model is not the requirement:
// multimodal_extractor@v2 runs google/gemini-embedding-2
// (3072-d) over a bucket, with no inference of your own.

Capabilities

Multimodal retrieval from image and text inputs
Generative matching objective for hard query-candidate pairs
Single-vector retrieval path with richer alignment than plain contrastive training
Apache 2.0 license