Gliclass Large V1.0 Init
GLiClass is an efficient zero-shot classifier trained on synthetic data, suitable for topic classification, sentiment analysis, and reranking tasks in RAG workflows.
Downloads 85
Release Time : 6/3/2024
Model Overview
A lightweight sequence classification model inspired by GLiNER, supporting zero-shot learning with computational efficiency comparable to cross-encoders while maintaining similar performance.
Model Features
Efficient zero-shot classification
Classification can be completed in a single forward pass, with computational efficiency superior to traditional cross-encoders
Multi-task applicability
Supports various text processing tasks including topic classification, sentiment analysis, and RAG reranking
Business-friendly
Trained on synthetic data, making it safe for commercial applications
Model Capabilities
Zero-shot text classification
Multi-label classification
Sentiment analysis
Retrieval-Augmented Generation (RAG) reranking
Use Cases
Content classification
News topic classification
Automatic multi-topic labeling for news texts
Achieves F1 score of 0.7516 on AG_NEWS dataset
Sentiment analysis
Review sentiment recognition
Identifies sentiment tendencies in user reviews
Achieves F1 score of 0.9404 on IMDB dataset
Information retrieval
RAG result reranking
Optimizes document ranking in Retrieval-Augmented Generation workflows
Featured Recommended AI Models