Arsh9210/Privasis-Cleaner-4B

Privasis-Cleaner-4B: Text Sanitization Model

Privasis-Cleaner-4B is a specialized 4 billion parameter language model, based on Qwen3 4B Instruct, engineered for text sanitization. Its core function is to identify and remove or abstract sensitive information from raw text according to explicit user instructions. This capability is crucial for maintaining data privacy and compliance.

Key Capabilities:

• Instruction-driven Sanitization: Users provide specific instructions (e.g., "Remove all person names, exact dates, and exact locations") to guide the model's sanitization process.
• Privacy-Preserving: Designed to help organizations and researchers handle sensitive text by redacting PII/PHI, enabling privacy-preserving research, and ensuring content compliance.
• Fine-tuned Performance: The model was fine-tuned on 37,000 instruction-input-output triplets, optimizing its ability to accurately sanitize text based on diverse requirements.
• Lightweight Architecture: Built on a 4B parameter Qwen3 model, offering efficient deployment for text cleaning tasks.

Ideal Use Cases:

• Automated Redaction: Automatically redacting Personally Identifiable Information (PII) or Protected Health Information (PHI) from documents.
• Compliance Pipelines: Integrating into workflows to meet regulatory requirements such as GDPR and HIPAA.
• Data Preprocessing: Preparing datasets for research or analysis by removing sensitive details while preserving contextual integrity.
• Content Moderation: Sanitizing user-generated content to remove unwanted sensitive data before publication or storage.

This model is intended for research and non-commercial use, as specified by its NVIDIA License (Non-Commercial).