Self-Guided Defense: Adaptive Safety Alignment for Reasoning Models via Synthesized Guidelines
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917184259751936 |
|---|---|
| author | Wang, Yuhang Zhu, Yanxu Lu, Dongyuan Sang, Jitao |
| author_facet | Wang, Yuhang Zhu, Yanxu Lu, Dongyuan Sang, Jitao |
| contents | Reasoning models have demonstrated remarkable capabilities in complex reasoning tasks. However, ensuring their safety against adversarial jailbreak prompts remains a critical challenge. Due to the covert and deceptive nature of such prompts, they can often evade built-in safety mechanisms and lead to the generation of harmful content. This underscores the need for an adaptive safety alignment approach that enables models to autonomously reinforce their defenses in response to adversarial inputs. This paper introduces the Synthesized Guideline-based Adaptive Safety Alignment (SGASA) framework, which internalizes model-generated safety guidelines to strengthen models' ability to enhance robustness against harmful adversarial prompts while minimizing unnecessary refusals of benign requests. SGASA consists of two key stages: Data Pre-synthesis, which generates safety guidelines and augmented prompts; and Alignment Fine-tuning, which leverages Supervised Fine-tuning (SFT) and Direct Preference Optimization (DPO) to embed these guidelines into the model. Extensive experiments across multiple datasets demonstrate that SGASA significantly improves model safety, validating its adaptive and scalable effectiveness. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_21214 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Self-Guided Defense: Adaptive Safety Alignment for Reasoning Models via Synthesized Guidelines Wang, Yuhang Zhu, Yanxu Lu, Dongyuan Sang, Jitao Computation and Language Artificial Intelligence Reasoning models have demonstrated remarkable capabilities in complex reasoning tasks. However, ensuring their safety against adversarial jailbreak prompts remains a critical challenge. Due to the covert and deceptive nature of such prompts, they can often evade built-in safety mechanisms and lead to the generation of harmful content. This underscores the need for an adaptive safety alignment approach that enables models to autonomously reinforce their defenses in response to adversarial inputs. This paper introduces the Synthesized Guideline-based Adaptive Safety Alignment (SGASA) framework, which internalizes model-generated safety guidelines to strengthen models' ability to enhance robustness against harmful adversarial prompts while minimizing unnecessary refusals of benign requests. SGASA consists of two key stages: Data Pre-synthesis, which generates safety guidelines and augmented prompts; and Alignment Fine-tuning, which leverages Supervised Fine-tuning (SFT) and Direct Preference Optimization (DPO) to embed these guidelines into the model. Extensive experiments across multiple datasets demonstrate that SGASA significantly improves model safety, validating its adaptive and scalable effectiveness. |
| title | Self-Guided Defense: Adaptive Safety Alignment for Reasoning Models via Synthesized Guidelines |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2511.21214 |