SaRO: Enhancing LLM Safety through Reasoning-based Alignment

• URL: http://arxiv.org/abs/2504.09420v1
• Date: Sun, 13 Apr 2025 03:36:06 GMT
• Title: SaRO: Enhancing LLM Safety through Reasoning-based Alignment
• Authors: Yutao Mou, Yuxiao Luo, Shikun Zhang, Wei Ye,
• Abstract summary: Current safety alignment techniques for large language models (LLMs) face two key challenges.<n>Under-generalization leaves models vulnerable to novel jailbreak attacks, and over-alignment leads to the excessive refusal of benign instructions.<n>We propose the Safety-oriented Reasoning Optimization Framework (SaRO) to incorporate safety-policy-driven reasoning into the alignment process.
• Score: 20.754670444745067
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Current safety alignment techniques for large language models (LLMs) face two key challenges: (1) under-generalization, which leaves models vulnerable to novel jailbreak attacks, and (2) over-alignment, which leads to the excessive refusal of benign instructions. Our preliminary investigation reveals semantic overlap between jailbreak/harmful queries and normal prompts in embedding space, suggesting that more effective safety alignment requires a deeper semantic understanding. This motivates us to incorporate safety-policy-driven reasoning into the alignment process. To this end, we propose the Safety-oriented Reasoning Optimization Framework (SaRO), which consists of two stages: (1) Reasoning-style Warmup (RW) that enables LLMs to internalize long-chain reasoning through supervised fine-tuning, and (2) Safety-oriented Reasoning Process Optimization (SRPO) that promotes safety reflection via direct preference optimization (DPO). Extensive experiments demonstrate the superiority of SaRO over traditional alignment methods.

Related papers

• ERPO: Advancing Safety Alignment via Ex-Ante Reasoning Preference Optimization [36.609297811592185]
  Ex-Ante Reasoning Preference Optimization (ERPO) is a novel safety alignment framework for large language models.<n>Our approach consists of three stages: first, equipping the model with Ex-Ante reasoning through supervised fine-tuning (SFT); second, enhancing safety, usefulness, and efficiency via Direct Preference Optimization (DPO); and third, mitigating inference latency with a length-controlled iterative preference optimization strategy.
  arXiv  Detail & Related papers  (2025-04-03T16:07:38Z)
• Safety is Not Only About Refusal: Reasoning-Enhanced Fine-tuning for Interpretable LLM Safety [31.933503076797148]
  Large Language Models (LLMs) are vulnerable to jailbreak attacks that exploit weaknesses in traditional safety alignment.<n>We propose Reasoning-enhanced Finetuning for interpretable LLM Safety (Rational)<n>Rational trains models to engage in explicit safe reasoning before response.
  arXiv  Detail & Related papers  (2025-03-06T22:47:45Z)
• Improving LLM Safety Alignment with Dual-Objective Optimization [65.41451412400609]
  Existing training-time safety alignment techniques for large language models (LLMs) remain vulnerable to jailbreak attacks.<n>We propose an improved safety alignment that disentangles DPO objectives into two components: (1) robust refusal training, which encourages refusal even when partial unsafe generations are produced, and (2) targeted unlearning of harmful knowledge.
  arXiv  Detail & Related papers  (2025-03-05T18:01:05Z)
• STAIR: Improving Safety Alignment with Introspective Reasoning [44.780098674618614]
  We propose STAIR, a framework that integrates SafeTy Alignment with Itrospective Reasoning.<n>We show that STAIR effectively mitigates harmful outputs while better preserving helpfulness, compared to instinctive alignment strategies.<n>With test-time scaling, STAIR achieves a safety performance comparable to Claude-3.5 against popular jailbreak attacks.
  arXiv  Detail & Related papers  (2025-02-04T15:02:55Z)
• Deliberative Alignment: Reasoning Enables Safer Language Models [64.60765108418062]
  We introduce Deliberative Alignment, a new paradigm that teaches the model safety specifications and trains it to explicitly recall and accurately reason over the specifications before answering.<n>We used this approach to align OpenAI's o-series models, and achieved highly precise adherence to OpenAI's safety policies, without requiring human-written chain-of-thoughts or answers.
  arXiv  Detail & Related papers  (2024-12-20T21:00:11Z)
• Superficial Safety Alignment Hypothesis [8.297367440457508]
  We propose the Superficial Safety Alignment Hypothesis (SSAH), which posits that safety alignment should teach an otherwise unsafe model to choose the correct reasoning direction. We identify four types of attribute-critical components in safety-aligned large language models (LLMs) Our findings show that freezing certain safety-critical components 7.5% during fine-tuning allows the model to retain its safety attributes while adapting to new tasks.
  arXiv  Detail & Related papers  (2024-10-07T19:53:35Z)
• SCANS: Mitigating the Exaggerated Safety for LLMs via Safety-Conscious Activation Steering [56.92068213969036]
  Safety alignment is indispensable for Large Language Models (LLMs) to defend threats from malicious instructions.<n>Recent researches reveal safety-aligned LLMs prone to reject benign queries due to the exaggerated safety issue.<n>We propose a Safety-Conscious Activation Steering (SCANS) method to mitigate the exaggerated safety concerns.
  arXiv  Detail & Related papers  (2024-08-21T10:01:34Z)
• One-Shot Safety Alignment for Large Language Models via Optimal Dualization [64.52223677468861]
  This paper presents a perspective of dualization that reduces constrained alignment to an equivalent unconstrained alignment problem. We do so by pre-optimizing a smooth and convex dual function that has a closed form. Our strategy leads to two practical algorithms in model-based and preference-based settings.
  arXiv  Detail & Related papers  (2024-05-29T22:12:52Z)
• ROSE Doesn't Do That: Boosting the Safety of Instruction-Tuned Large Language Models with Reverse Prompt Contrastive Decoding [89.0074567748505]
  We present reverse prompt contrastive decoding (ROSE), a simple-yet-effective method to boost the safety of existing instruction-tuned LLMs without any additional training. Experiments on 6 safety and 2 general-purpose tasks show that, our ROSE not only brings consistent and significant safety improvements (up to +13.8% safety score) upon 5 types of instruction-tuned LLMs, but also benefits the general-purpose ability of LLMs.
  arXiv  Detail & Related papers  (2024-02-19T06:58:42Z)
• On Prompt-Driven Safeguarding for Large Language Models [172.13943777203377]
  We find that in the representation space, the input queries are typically moved by safety prompts in a "higher-refusal" direction. Inspired by these findings, we propose a method for safety prompt optimization, namely DRO. Treating a safety prompt as continuous, trainable embeddings, DRO learns to move the queries' representations along or opposite the refusal direction, depending on their harmfulness.
  arXiv  Detail & Related papers  (2024-01-31T17:28:24Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.