H

Has 820m

Developed by SecurityXuanwuLab
A privacy protection model developed by Tencent Security Xuanwu Lab, safeguarding user privacy by hiding sensitive information and restoring output content.
Downloads 2,730
Release Time : 11/18/2023

Model Overview

The Hide-and-Seek model achieves privacy protection through two subtasks: 'hide' replaces sensitive entities with random ones (privacy anonymization), while 'seek' restores the replaced content in outputs to original text (information recovery).

Model Features

Dual Privacy Protection Mechanism
Implements a complete privacy protection workflow through the two phases of 'hide' anonymization and 'seek' restoration
Efficient Inference
Achieves high-speed inference on mobile devices with NF4 quantization (180-200 tokens/sec on MacBook M2)
Multi-task Support
Natively supports text polishing, summarization, translation, reading comprehension, and text classification tasks
Zero-shot Extension
Supports custom task extensions in Zero Shot mode

Model Capabilities

Privacy anonymization processing
Information restoration processing
Text rewriting
Summarization generation
Text translation
Reading comprehension
Text classification

Use Cases

Privacy-preserving text processing
Sensitive Information Anonymization
Replaces personally identifiable information and trade secrets in text with random entities
Protects original information from being accessed by third-party services
Privacy-preserving Translation
Processes translation after anonymization and restores sensitive information afterward
Enables translation services without exposing original sensitive content
Content Generation
Privacy-preserving Summarization
Generates summaries for texts containing sensitive information while protecting original content
Produces summaries free of sensitive information
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase