Preventing Safety Drift in Large Language Models via Coupled Weight and Activation Constraints

arXiv:2604.12384v1 Announce Type: new Abstract: Safety alignment in Large Language Models (LLMs) remains highly fragile during fine-tuning, where even benign adaptation can degrade pre-trained refusal behaviors and enable harmful responses. Existing defenses typically constrain either weights or activations in isolation, without considering their coupled effects on safety. In this paper, we first theoretically demonstrate that constraining either weights or activations alone is insufficient for safety preservation. To robustly preserve safety alignment, we propose Coupled Weight and Activation Constraints (CWAC), a novel approach that simultaneously enforces a precomputed safety subspace on weight updates and applies targeted regularization to safety-critical features identified by sparse autoencoders. Extensive experiments across four widely used LLMs and diverse downstream tasks show that CWAC consistently achieves the lowest harmful scores with minimal impact on fine-tuning accuracy, substantially outperforming strong baselines even under high harmful data ratios.

Related

• Deliberative Alignment is Deep, but Uncertainty Remains: Inference time safety improvement in reasoning via attribution of unsafe behavior to base model
• ASGuard: Activation-Scaling Guard to Mitigate Targeted Jailbreaking Attack
• Large Language Models Generate Harmful Content Using a Distinct, Unified Mechanism
• The Art of (Mis)alignment: How Fine-Tuning Methods Effectively Misalign and Realign LLMs in Post-Training

Source: arXiv cs.AI | 2026-04-15