PRUNE is a novel approach that enables sequential editing of large language models without degrading their general capabilities.

• Addresses the critical trade-off between knowledge updates and performance degradation
• Implements a perturbation-restrained mechanism that maintains model integrity during edits
• Theoretically analyzes the factors affecting model performance during sequential editing
• Demonstrates superior results compared to existing model editing techniques

For Education, this research enables more efficient knowledge updates in educational AI systems, allowing for curriculum adjustments without costly retraining while maintaining overall system performance.

Perturbation-Restrained Sequential Model Editing