Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2602.00038 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910006478110720 |
|---|---|
| author | Zhou, Guanghao Qiu, Panjia Chen, Cen Li, Hongyu Chu, Mingyuan Zhang, Xin Zhou, Jun |
| author_facet | Zhou, Guanghao Qiu, Panjia Chen, Cen Li, Hongyu Chu, Mingyuan Zhang, Xin Zhou, Jun |
| contents | The safety mechanisms of large language models (LLMs) exhibit notable fragility, as even fine-tuning on datasets without harmful content may still undermine their safety capabilities. Meanwhile, existing safety alignment methods predominantly rely on the fine-tuning process, which inadvertently leads to the increased complexity and computational resources required. To address these issues, we introduce LSSF, a novel safety re-alignment framework with \underline{L}ow-Rank \underline{S}afety \underline{S}ubspace \underline{F}usion. Our proposed method exploits the low-rank characteristics of safety information in LLMs by constructing a low-rank projection matrix to extract the principal components of safety vectors. Notably, this projection matrix represents the low-rank safety subspace of the LLMs, which we have observed to remain stable during fine-tuning process and is isolated from the model's general capabilities. These principal components are used to effectively restore safety alignment when combined with fine-tuned LLMs through linear arithmetic. Additionally, to account for the varying encoding densities of safety information across different layers of LLMs, we propose a novel metric called safety singular value entropy. This metric quantifies the encoding density and allows for the dynamic computation of the safety-critical rank for each safety vector. Extensive experiments demonstrate that our proposed post-hoc alignment method can effectively restore the safety alignment of fine-tuned models with minimal impact on their performance in downstream tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_00038 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | LSSF: Safety Alignment for Large Language Models through Low-Rank Safety Subspace Fusion Zhou, Guanghao Qiu, Panjia Chen, Cen Li, Hongyu Chu, Mingyuan Zhang, Xin Zhou, Jun Computers and Society Artificial Intelligence The safety mechanisms of large language models (LLMs) exhibit notable fragility, as even fine-tuning on datasets without harmful content may still undermine their safety capabilities. Meanwhile, existing safety alignment methods predominantly rely on the fine-tuning process, which inadvertently leads to the increased complexity and computational resources required. To address these issues, we introduce LSSF, a novel safety re-alignment framework with \underline{L}ow-Rank \underline{S}afety \underline{S}ubspace \underline{F}usion. Our proposed method exploits the low-rank characteristics of safety information in LLMs by constructing a low-rank projection matrix to extract the principal components of safety vectors. Notably, this projection matrix represents the low-rank safety subspace of the LLMs, which we have observed to remain stable during fine-tuning process and is isolated from the model's general capabilities. These principal components are used to effectively restore safety alignment when combined with fine-tuned LLMs through linear arithmetic. Additionally, to account for the varying encoding densities of safety information across different layers of LLMs, we propose a novel metric called safety singular value entropy. This metric quantifies the encoding density and allows for the dynamic computation of the safety-critical rank for each safety vector. Extensive experiments demonstrate that our proposed post-hoc alignment method can effectively restore the safety alignment of fine-tuned models with minimal impact on their performance in downstream tasks. |
| title | LSSF: Safety Alignment for Large Language Models through Low-Rank Safety Subspace Fusion |
| topic | Computers and Society Artificial Intelligence |
| url | https://arxiv.org/abs/2602.00038 |