R

Reranker ModernBERT Large Gooaq Bce

Developed by tomaarsen
This is a cross-encoder model fine-tuned from ModernBERT-large, used to calculate the scores of text pairs, suitable for text re-ranking and semantic search tasks.
Downloads 596
Release Time : 3/25/2025

Model Overview

This model is based on the ModernBERT-large architecture and trained through the sentence-transformers library. It is specifically designed for similarity scoring and re-ranking tasks of text pairs. It performs well on multiple datasets and is particularly good at ranking content related to questions and answers.

Model Features

High-performance re-ranking
Achieved an NDCG@10 score of 0.7946 on the GooAQ development set, significantly outperforming the baseline model
Large context support
Supports a maximum sequence length of 8192 tokens, suitable for processing long texts
Multi-dataset adaptability
Performs well on multiple datasets such as NanoMSMARCO, NanoNFCorpus, and NanoNQ

Model Capabilities

Text similarity scoring
Search result re-ranking
Question-answer pair relevance assessment
Semantic search enhancement

Use Cases

Search engine optimization
Search result re-ranking
Re-rank the results returned by the search engine to improve the ranking of the most relevant results
NDCG@10 increased by 20.34% on the GooAQ dataset
Question-answering system
Answer relevance assessment
Evaluate the relevance of candidate answers to the question and filter out the best answer
MAP reached 0.6103 on the NanoNQ dataset
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Ā© 2025AIbase