TAROT introduces a novel approach to disguising authorship while preserving text utility through reinforcement learning techniques.

• Balances privacy protection and content utility more effectively than previous methods
• Uses policy optimization to create text transformations that hide author identity
• Achieves task-specific obfuscation that maintains the original purpose of the text
• Demonstrates enhanced resistance to de-anonymization attacks

This advancement is critical for security applications where protecting user identity is paramount while maintaining functional communication - particularly important as language models make authorship attribution increasingly accessible.

TAROT: Task-Oriented Authorship Obfuscation Using Policy Optimization Methods