J

Jina Reranker V2 Base Multilingual Wiki Tr Rag Prefix

Developed by SMARTICT
A fine-tuned cross-encoder model based on jina-reranker-v2-base-multilingual for text reranking and semantic search
Downloads 173
Release Time : 4/15/2025

Model Overview

This is a cross-encoder model fine-tuned from jina-reranker-v2-base-multilingual, capable of computing matching scores for text pairs, primarily used for text reranking and semantic search tasks.

Model Features

Multilingual Support
Based on a multilingual foundation model, supporting text reranking in multiple languages
High-Performance Reranking
Excellent performance on multiple evaluation datasets, especially achieving 0.9386 nDCG@10 on the gooaq development set
Long-Text Processing
Supports sequences up to 1024 tokens, suitable for processing longer texts

Model Capabilities

Text Pair Matching Scoring
Semantic Search Reranking
Multilingual Text Processing

Use Cases

Information Retrieval
QA System Reranking
Rerank candidate answers in QA systems to improve the ranking of correct answers
Achieved 0.6937 average precision on the NanoNQ dataset
Document Retrieval
Rerank documents returned by search engines to improve relevance
Achieved 0.5847 average precision on the NanoMSMARCO dataset
Recommendation Systems
Content Recommendation
Rank recommended content by relevance to improve recommendation quality
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase