This research introduces a novel two-stage methodology for selectively removing sensitive information from large language models without compromising their overall capabilities.

• Combines causal mediation analysis with layer-specific optimization
• Enables targeted removal of specific data associations
• Maintains model performance on general tasks
• Addresses critical privacy and security concerns for public AI deployment

This advancement matters for security professionals as it provides a practical solution to one of the major barriers to safe AI deployment - the ability to selectively "forget" sensitive data without degrading model utility.

Original Paper: SHA256 at SemEval-2025 Task 4: Selective Amnesia -- Constrained Unlearning for Large Language Models via Knowledge Isolation