[Paper] CNT: Safety-oriented Function Reuse across LLMs via Cross-Model Neuron Transfer
Source: arXiv - 2603.18449v1
Overview
The widespread deployment of large language models (LLMs) calls for post-hoc methods that can flexibly adapt models to evolving safety requirements. Meanwhile, the rapidly expanding open-source LLM ecosystem has produced a diverse collection of models that already exhibit various safety-related functionalities. This motivates a shift from constructing safety functionality from scratch to reusing existing functionality from external models, thereby avoiding costly data collection and training procedures. In this paper, we present Cross-Model Neuron Transfer (CNT), a post-hoc method that reuses safety-oriented functionality by transferring a minimal subset of neurons from an open-source donor LLM to a target LLM. By operating at the neuron level, CNT enables modular function-level adaptation, supporting both function addition andfunction deletion. We evaluate CNT on seven popular LLMs across three representative applications: safety disalignment, alignment enhancement, and bias removal. Experimental results show that CNT achieves targeted safety-oriented functionality transfer with minimal performance degradation (less than 1% for most models), consistently outperforming five baselines, demonstrating its generality and practical effectiveness.
Key Contributions
This paper presents research in the following areas:
- cs.CR
- cs.SE
Methodology
Please refer to the full paper for detailed methodology.
Practical Implications
This research contributes to the advancement of cs.CR.
Authors
- Yue Zhao
- Yujia Gong
- Ruigang Liang
- Shenchen Zhu
- Kai Chen
- Xuejing Yuan
- Wangjun Zhang
Paper Information
- arXiv ID: 2603.18449v1
- Categories: cs.CR, cs.SE
- Published: March 19, 2026
- PDF: Download PDF