B

Bge Reranker V2 M3 Ko

Developed by dragonkue
This is a Korean reranking model optimized based on BAAI/bge-reranker-v2-m3, primarily used for text reranking tasks.
Downloads 877
Release Time : 10/16/2024

Model Overview

This model is a cross-encoder that directly takes a query and document as input and outputs a similarity score. By inputting queries and passages, the model returns relevance scores, suitable for information retrieval and document reranking tasks.

Model Features

Multilingual support
Supports Korean and English, with special optimization for Korean.
High-precision reranking
Directly computes similarity scores for text pairs, achieving higher accuracy than dual-encoder models.
Multiple usage options
Supports usage via Transformers, SentenceTransformers, and FlagEmbedding libraries.

Model Capabilities

Text similarity calculation
Document reranking
Information retrieval

Use Cases

Information retrieval
Financial document retrieval
Used for retrieving Korean financial documents such as legal provisions and policy documents.
Achieved a Top-1 F1 score of 0.9123 in Korean financial domain benchmarks.
Question answering systems
Question-answer matching
Used to calculate the relevance between questions and candidate answers, selecting the best-matching answer.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase