R

Ruri V3 Reranker 310m

Developed by cl-nagoya
A Japanese general-purpose reranking model built on ModernBERT-Ja, featuring top-tier performance and long-sequence processing capabilities
Downloads 1,100
Release Time : 4/6/2025

Model Overview

Ruri v3 Reranker is a Japanese general-purpose reranking model based on the ModernBERT-Ja architecture, supporting sequence processing of up to 8192 tokens and excelling across multiple domains.

Model Features

Top-Tier Performance
Excels in multi-domain data with strong robustness
Long-Sequence Processing
Supports sequence processing of up to 8192 tokens
Extended Vocabulary
100K token extended vocabulary (v1/v2 versions had 32K)
Efficient Attention Mechanism
Incorporates FlashAttention technology for improved computational efficiency
Pure SentencePiece Tokenizer
Uses a pure SentencePiece tokenizer for more efficient Japanese text processing

Model Capabilities

Japanese text reranking
Cross-document relevance scoring
Long text processing
Multi-domain text understanding

Use Cases

Information Retrieval
Search Result Reranking
Reorders search engine results by relevance
Improves search result relevance and accuracy
Question Answering Systems
Answer Candidate Ranking
Ranks multiple candidate answers generated by a QA system
Selects the most relevant answer to display to users
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase