M

Modernbert Base Msmarco

Developed by joe32140
This model is a fine-tuned sentence embedding model based on ModernBERT-base, specifically designed for sentence similarity tasks and supports English text processing.
Downloads 4,695
Release Time : 12/20/2024

Model Overview

By fine-tuning the ModernBERT-base architecture, this model focuses on sentence similarity calculation and feature extraction tasks, suitable for scenarios like information retrieval and question-answer matching.

Model Features

Efficient sentence embeddings
Generates high-quality sentence embeddings through fine-tuning ModernBERT-base, suitable for similarity calculation and information retrieval.
Optimized loss function
Trained using CachedMultipleNegativesRankingLoss, enhancing performance in negative sample contrastive learning.
Large-scale dataset training
Trained on the msmarco-co-condenser-margin-mse-sym-mnrl-mean-v1 dataset (approximately 11.66 million data points), demonstrating strong generalization capabilities.

Model Capabilities

Sentence similarity calculation
Feature extraction
Information retrieval
Question-answer matching

Use Cases

Information retrieval
Question-answer matching
Finds the most relevant answer from candidate responses based on user queries.
Achieves a cosine accuracy of 0.984 on the msmarco co condenser development set.
Text classification
Amazon review classification
Classifies Amazon product reviews.
Achieves an accuracy of 32.318 on the MTEB Amazon review classification (English) task.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase