Q-realign: Piggybacking Realignment on Quantization for Safe and Efficient LLM Deployment
This addresses safety risks in LLM deployment for practitioners, offering a practical solution that decouples safety from fine-tuning, though it is incremental as it builds on existing quantization techniques.
The paper tackles the problem of safety alignment erosion during task-specific fine-tuning of large language models (LLMs) by proposing Q-realign, a post-hoc defense method that reduces unsafe behaviors while preserving task performance, with notable efficiency gains such as recovering safety alignment on a 7B LLM within 40 minutes on a single RTX 4090.
Public large language models (LLMs) are typically safety-aligned during pretraining, yet task-specific fine-tuning required for deployment often erodes this alignment and introduces safety risks. Existing defenses either embed safety recovery into fine-tuning or rely on fine-tuning-derived priors for post-hoc correction, leaving safety recovery tightly coupled with training and incurring high computational overhead and a complex workflow. To address these challenges, we propose \texttt{Q-realign}, a post-hoc defense method based on post-training quantization, guided by an analysis of representational structure. By reframing quantization as a dual-objective procedure for compression and safety, \texttt{Q-realign} decouples safety alignment from fine-tuning and naturally piggybacks into modern deployment pipelines. Experiments across multiple models and datasets demonstrate that our method substantially reduces unsafe behaviors while preserving task performance, with significant reductions in memory usage and GPU hours. Notably, our approach can recover the safety alignment of a fine-tuned 7B LLM on a single RTX 4090 within 40 minutes. Overall, our work provides a practical, turnkey solution for safety-aware deployment.