mxbai rerank large v1

Introduction

The MXBAI-RERANK-LARGE-V1 model is part of a powerful reranker family developed by Mixedbread. It is designed to enhance search capabilities and is highly effective when combined with keyword search methods. The model can outperform semantic search systems under various conditions.

Architecture

The MXBAI-RERANK-LARGE-V1 model employs the latest in transformer-based architectures, leveraging the capabilities of sentence-transformers. It supports various libraries, including transformers.js, and is optimized for English language text classification tasks.

Training

The model performs exceptionally well, achieving high scores in metrics such as NDCG@10 and Accuracy@3. Evaluations were conducted using Pyserini on 11 datasets from BEIR, showcasing its effectiveness against both lexical and semantic search baselines.

Guide: Running Locally

1. Setup Environment: Ensure you have Python and npm installed for JavaScript usage.
2. Install Dependencies:
   • For Python:

     pip install -U sentence-transformers
     

   • For JavaScript:

     npm i @xenova/transformers
     

3. Load the Model:
   • Python:

     from sentence_transformers import CrossEncoder
     model = CrossEncoder("mixedbread-ai/mxbai-rerank-large-v1")
     

   • JavaScript:

     import { AutoTokenizer, AutoModelForSequenceClassification } from '@xenova/transformers';
     const model = await AutoModelForSequenceClassification.from_pretrained('mixedbread-ai/mxbai-rerank-large-v1');
     

4. Perform Reranking: Use the model to rerank documents based on a query.

For substantial computational tasks, consider using cloud GPUs from providers like AWS or Google Cloud for enhanced performance.

License

The MXBAI-RERANK-LARGE-V1 model is licensed under Apache License 2.0, allowing for both commercial and non-commercial use while ensuring attribution to the original creators.