Surgical Knowledge Rewrite in Compact LLMs: An 'Unlearn-then-Learn' Strategy with (IA³) for Localized Factual Modulation and Catastrophic Forgetting Mitigation

2025/08/09 by Ngugi, Stanley
#Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · doi:10.48550/arxiv.2508.07075

Abstract

Large Language Models (LLMs) struggle with dynamic knowledge updates, especially when new information conflicts with deeply embedded facts. Such conflicting factual edits often lead to two critical issues: resistance to adopting the new fact and severe catastrophic forgetting of unrelated knowledge. This paper introduces and evaluates a novel "unlearn-then-learn" strategy for precise knowledge editing in LLMs, leveraging the parameter-efficient fine-tuning (PEFT) technique, Infused Adapter by Inhibiting and Amplifying Inner Activations (IA³). Crucially, this two-stage approach is powered by an initial circuit localization phase that identifies and targets the specific internal components responsible for encoding the conflicting fact. Through a rigorous experimental methodology on microsoft/Phi-3-mini-4k-instruct, we demonstrate that this mechanistically informed two-stage approach achieves near-perfect accuracy (98.50%) for the new, modulated fact while simultaneously effectively suppressing the original conflicting fact (96.00% forget rate). Critically, our strategy exhibits unprecedented localization (72.00% F_control accuracy), dramatically mitigating catastrophic forgetting observed in direct fine-tuning approaches (which showed as low as ~20% F_control accuracy), a direct benefit of our targeted interpretability-guided intervention. Furthermore, qualitative analysis reveals a nuanced mechanism of "soft forgetting," where original knowledge is suppressed from default retrieval but remains latent and conditionally accessible, enhancing model safety and control. These findings represent a significant advancement towards precise, localized, and safe knowledge management in compact LLMs.

Citations

• Uncertainty Quantification in Large Language Models Through Convex Hull Analysis
• QLoRA: Efficient Finetuning of Quantized LLMs
• Mass-Editing Memory in a Transformer
• LoRA: Low-Rank Adaptation of Large Language Models

Related

• Precise Localization of Memories: A Fine-grained Neuron-level Knowledge Editing Technique for LLMs
• LoKI: Low-damage Knowledge Implanting of Large Language Models
• SEAT: Sparse Entity-Aware Tuning for Knowledge Adaptation while Preserving Epistemic Abstention
• LLM Surgery: Efficient Knowledge Unlearning and Editing in Large Language Models
• Balancing Knowledge Updates: Toward Unified Modular Editing in LLMs