Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Enhancing Safety of Large Language Models via Embedding Space Separation

About

Large language models (LLMs) have achieved impressive capabilities, yet ensuring their safety against harmful prompts remains a critical challenge. Recent work has revealed that the latent representations (embeddings) of harmful and safe queries in LLMs typically exhibit linear separability, a property that has been exploited to construct attacks by perturbing the embeddings of harmful queries towards the safe subspace. Motivated by this observation, we propose a representation-level fine-tuning approach, named Embedding Space Separation (ES2), which improves LLM safety by explicitly enlarging the distance between harmful and safe representations in the embedding space. To prevent degradation of model's general capabilities, we introduce a Kullback-Leibler (KL) divergence regularization term into the loss function, which constrains the logits of the fine-tuned model to align with those of the original base model on harmless inputs. We evaluate our method on several open-source LLMs using standard safety benchmarks. Extensive experimental results demonstrate that our approach substantially improves model safety while maintaining comparable general capabilities.

Xu Zhao, Xiting Wang, Weiran Shen• 2026

Related benchmarks

TaskDatasetResultRank
Defense RobustnessRepE Attack
DSR Keyword Match Rate100
16
Defense RobustnessSoft Prompt Attack
DSR Keyword Success Rate100
16
Defense RobustnessSCAV Attack
DSR (Keyword)80
16
Defense RobustnessAutoDAN Attack
Keyword Success Rate100
16
Defense RobustnessGCG Attack
Keyword Match Rate100
16
Defense RobustnessDirect Inquiry Vanilla
Keyword Match Rate100
16
Showing 6 of 6 rows

Other info

Follow for update