D

Deberta V3 Large Zeroshot V2.0 C

Developed by MoritzLaurer
A DeBERTa-v3-large model specifically designed for efficient zero-shot classification, trained on fully commercially friendly synthetic data and NLI datasets, supporting GPU/CPU inference
Downloads 1,560
Release Time : 3/20/2024

Model Overview

A zero-shot classification model based on the DeBERTa-v3-large architecture, achieving text classification without training data through Natural Language Inference (NLI) task formatting, suitable for multi-domain scenarios

Model Features

Commercially Friendly Data
Trained on synthetic data generated by Mixtral and commercially friendly datasets like MNLI/FEVER-NLI, meeting strict licensing requirements
Zero-shot Classification
Performs text classification tasks without training data by converting any classification task into NLI format using hypothesis templates
High-performance Architecture
Based on the DeBERTa-v3-large architecture, achieving an average F1 score of 0.676 across 28 text classification tasks, outperforming comparable benchmark models
Flexible Templates
Supports custom hypothesis templates, similar to LLM prompt engineering, to optimize classification performance

Model Capabilities

Zero-shot text classification
Multi-class classification (single-label/multi-label)
Cross-domain classification (supports 25+ industries)

Use Cases

Content Classification
News Topic Classification
Automatically categorizes news into topics such as politics, economy, entertainment, etc.
Shows high accuracy in synthetic data testing
Social Media Content Moderation
Identifies categories of prohibited content (hate speech, misinformation, etc.)
Business Analysis
Customer Feedback Classification
Automatically categorizes user reviews into dimensions like product features, service quality, etc.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase