BAAI/bge-reranker-v2-m3
🧠 AI ModelBAAI
Multilingual cross-encoder reranker for 100+ languages.
The BGE-Reranker-v2-M3 is a cross-encoder reranker that takes a query-document pair as input and outputs a relevance score. It is built on the XLM-RoBERTa architecture, supporting over 100 languages. The model was fine-tuned on the M3-Embedding dataset, which includes massive, multilingual, and multi-task data. It outperforms previous rerankers like Cohere-rerank and multilingual versions of MonoT5 on benchmarks such as BeIR and MIRACL. The model uses a lightweight architecture with 568M parameters, making it efficient for production deployment. It is compatible with sentence-transformers and integrates seamlessly with frameworks like Faiss and Elasticsearch.
💡Highlights
- ├─Multilingual support for 100+ languages
- ├─Cross-encoder architecture for high accuracy
- └─568M parameters, efficient for deployment
🎯For
- ├─NLP researchers
- ├─Search engineers
- └─RAG developers