English

MoGU V2: Toward a Higher Pareto Frontier Between Model Usability and Security

Computation and Language 2025-09-09 v1

Abstract

As Large Language Models (LLMs) increasingly permeate human life, their security has emerged as a critical concern, particularly their ability to maintain harmless responses to malicious instructions. Although extensive methods have improved LLMs' security, they often lead to conservative, rejection-oriented responses that compromise practical usability. This presents a key challenge: how to advance the Pareto frontier between LLMs' usability and security, rather than necessitate a trade-off between them. To address this, we propose the MoGU framework, in which the intra-layer router dynamically allocates weights by sensing hidden states, thereby balancing the contributions of security-optimized and usability-optimized variants. Despite its initial potential, the MoGU framework faces limitations such as parameter redundancy and performance bottlenecks. To overcome these, we further propose an improved MoGU_v2 framework that establishes a tighter coupling between the routers and hidden states. In MoGU_v2, routers are embedded only in layers encoding highly classifiable security features, and backbone modules are activated during router optimization to enable bidirectional adaptation. MoGU_V2 exhibits strong adaptability and stable improvements across various series of LLMs, including mainstream LLMs serving as brains in various applications, on-device LLMs optimized for resource-constrained scenarios, and reasoning LLMs tailored for user interpretability. Meanwhile, even facing risks introduced by Instruction Fine-tuning, MoGU_v2 can easily restore security without compromising the task performance gains via a simple data-mix strategy. These comprehensive improvements highlight MoGU_V2 as a robust and versatile solution for mitigating security risks in real-world applications.

Keywords

Cite

@article{arxiv.2509.06807,
  title  = {MoGU V2: Toward a Higher Pareto Frontier Between Model Usability and Security},
  author = {Yanrui Du and Fenglei Fan and Sendong Zhao and Jiawei Cao and Ting Liu and Bing Qin},
  journal= {arXiv preprint arXiv:2509.06807},
  year   = {2025}
}
R2 v1 2026-07-01T05:26:40.797Z