M

Moritzlaurer Mdeberta V3 Base Mnli Xnli

Developed by MaagDeveloper
Natural language inference model supporting 100 languages, suitable for multilingual zero-shot classification tasks
Downloads 47
Release Time : 5/11/2025

Model Overview

Multilingual text classification model based on mDeBERTa-v3-base, fine-tuned on XNLI and MNLI datasets, excelling in zero-shot classification and natural language inference tasks

Model Features

Multilingual support
Supports zero-shot classification and natural language inference in 100 languages
High performance
As of December 2021, it was the best-performing multilingual base-scale Transformer model
Cross-lingual transfer capability
Can handle 85 languages involved in pre-training even without specific language training data

Model Capabilities

Multilingual text classification
Natural language inference
Zero-shot learning
Cross-lingual text understanding

Use Cases

Content classification
News classification
Automatically classify multilingual news into predefined categories (e.g., politics, economy)
Achieved an average accuracy of 80.8% on the XNLI test set
Semantic analysis
Textual entailment judgment
Determine the logical relationship between two texts (entailment/neutral/contradiction)
81.16% accuracy on Chinese test set
Featured Recommended AI Models
ยฉ 2025AIbase