GTE-Rerank-V2 is a second-generation text reranking model developed by Alibaba’s Tongyi Lab (part of the GTE family). It is designed to reorder a list of candidate passages by their relevance to a query, typically used after an initial retrieval step.
- Core features: Cross-encoder architecture for precise query–passage interaction; supports long inputs (up to 8,192 tokens); multilingual (English, Chinese, and others); improved accuracy over GTE-Rerank-V1; available via Hugging Face and ModelScope.
- Use cases: Two-stage retrieval pipelines (e.g., RAG, search engines, question answering), where a fast retriever (like GTE embeddings) fetches top‑k documents and GTE-Rerank-V2 reorders them for final ranking.
