Z

Zero Shot Classify SSTuning Base

Developed by DAMO-NLP-SG
This model is trained using self-supervised tuning methods and is suitable for zero-shot text classification tasks. It can perform sentiment analysis, topic classification, and other tasks without additional fine-tuning.
Downloads 44
Release Time : 5/18/2023

Model Overview

A zero-shot text classification model trained using the Self-Supervised Tuning (SSTuning) method, based on the RoBERTa-base architecture, supporting classification tasks with 2-20 labels.

Model Features

Self-supervised tuning
Uses First Sentence Prediction (FSP) as the learning objective, leveraging unlabeled data for tuning, enabling training without labeled data.
Zero-shot capability
Can be directly applied to new classification tasks without additional fine-tuning, supporting classification with 2-20 labels.
Multiple model variants
Offers model variants of different scales and performance, including base, large, ALBERT, and multilingual versions.

Model Capabilities

Zero-shot text classification
Sentiment analysis
Topic classification

Use Cases

Sentiment analysis
Review sentiment classification
Classify user reviews into positive or negative sentiment categories.
Confidence can reach up to 99.92%
Topic classification
News classification
Categorize news articles into predefined topic categories.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase