What Matters For Safety Alignment?

• URL: http://arxiv.org/abs/2601.03868v1
• Date: Wed, 07 Jan 2026 12:31:52 GMT
• Title: What Matters For Safety Alignment?
• Authors: Xing Li, Hui-Ling Zhen, Lihao Yin, Xianzhi Yu, Zhenhua Dong, Mingxuan Yuan,
• Abstract summary: This paper presents a comprehensive empirical study on the safety alignment capabilities of AI systems.<n>We systematically investigate and compare the influence of six critical intrinsic model characteristics and three external attack techniques.<n>We identify the LRMs GPT-OSS-20B, Qwen3-Next-80B-A3B-Thinking, and GPT-OSS-120B as the top-three safest models.
• Score: 38.86339753409445
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: This paper presents a comprehensive empirical study on the safety alignment capabilities. We evaluate what matters for safety alignment in LLMs and LRMs to provide essential insights for developing more secure and reliable AI systems. We systematically investigate and compare the influence of six critical intrinsic model characteristics and three external attack techniques. Our large-scale evaluation is conducted using 32 recent, popular LLMs and LRMs across thirteen distinct model families, spanning a parameter scale from 3B to 235B. The assessment leverages five established safety datasets and probes model vulnerabilities with 56 jailbreak techniques and four CoT attack strategies, resulting in 4.6M API calls. Our key empirical findings are fourfold. First, we identify the LRMs GPT-OSS-20B, Qwen3-Next-80B-A3B-Thinking, and GPT-OSS-120B as the top-three safest models, which substantiates the significant advantage of integrated reasoning and self-reflection mechanisms for robust safety alignment. Second, post-training and knowledge distillation may lead to a systematic degradation of safety alignment. We thus argue that safety must be treated as an explicit constraint or a core optimization objective during these stages, not merely subordinated to the pursuit of general capability. Third, we reveal a pronounced vulnerability: employing a CoT attack via a response prefix can elevate the attack success rate by 3.34x on average and from 0.6% to 96.3% for Seed-OSS-36B-Instruct. This critical finding underscores the safety risks inherent in text-completion interfaces and features that allow user-defined response prefixes in LLM services, highlighting an urgent need for architectural and deployment safeguards. Fourth, roleplay, prompt injection, and gradient-based search for adversarial prompts are the predominant methodologies for eliciting unaligned behaviors in modern models.

Related papers

• ProGuard: Towards Proactive Multimodal Safeguard [48.89789547707647]
  ProGuard is a vision-language proactive guard that identifies and describes out-of-distribution (OOD) safety risks.<n>We first construct a modality-balanced dataset of 87K samples, each annotated with both binary safety labels and risk categories.<n>We then train our vision-language base model purely through reinforcement learning to achieve efficient and concise reasoning.
  arXiv  Detail & Related papers  (2025-12-29T16:13:23Z)
• Quantifying CBRN Risk in Frontier Models [0.0]
  Frontier Large Language Models (LLMs) pose unprecedented dual-use risks through the potential proliferation of chemical, biological, radiological, and nuclear (CBRN) weapons knowledge.<n>We present the first comprehensive evaluation of 10 leading commercial LLMs against a novel CBRN dataset and a 180-prompt subset of the FORTRESS benchmark.
  arXiv  Detail & Related papers  (2025-10-24T03:55:24Z)
• SaFeR-VLM: Toward Safety-aware Fine-grained Reasoning in Multimodal Models [66.71948519280669]
  Multimodal Large Reasoning Models (MLRMs) demonstrate impressive crossmodal reasoning but often amplify safety risks under adversarial prompts.<n> Existing defenses mainly act at the output level and do not constrain the reasoning process, leaving models to implicit risks.<n>We propose SaFeR-VLM, which integrates four components and supports dynamic and interpretable safety decisions beyond surface-level filtering.
  arXiv  Detail & Related papers  (2025-10-08T10:39:12Z)
• DiffuGuard: How Intrinsic Safety is Lost and Found in Diffusion Large Language Models [50.21378052667732]
  We conduct an in-depth analysis of dLLM vulnerabilities to jailbreak attacks across two distinct dimensions: intra-step and inter-step dynamics.<n>We propose DiffuGuard, a training-free defense framework that addresses vulnerabilities through a dual-stage approach.
  arXiv  Detail & Related papers  (2025-09-29T05:17:10Z)
• Circumventing Safety Alignment in Large Language Models Through Embedding Space Toxicity Attenuation [13.971909819796762]
  Large Language Models (LLMs) have achieved remarkable success across domains such as healthcare, education, and cybersecurity.<n>Embedding space poisoning is a subtle attack vector where adversaries manipulate the internal semantic representations of input data to bypass safety alignment mechanisms.<n>We propose ETTA, a novel framework that identifies and attenuates toxicity-sensitive dimensions in embedding space via linear transformations.
  arXiv  Detail & Related papers  (2025-07-08T03:01:00Z)
• Analysing Safety Risks in LLMs Fine-Tuned with Pseudo-Malicious Cyber Security Data [2.549390156222399]
  Large language models (LLMs) have been used in many application domains, including cyber security.<n>Recent findings show that fine-tuning LLMs with pseudo-malicious cyber security data significantly compromises their safety.<n>This paper presents a comprehensive validation and extension of these safety risks using a different evaluation framework.
  arXiv  Detail & Related papers  (2025-05-15T05:22:53Z)
• SafeMLRM: Demystifying Safety in Multi-modal Large Reasoning Models [50.34706204154244]
  Acquiring reasoning capabilities catastrophically degrades inherited safety alignment.<n>Certain scenarios suffer 25 times higher attack rates.<n>Despite tight reasoning-answer safety coupling, MLRMs demonstrate nascent self-correction.
  arXiv  Detail & Related papers  (2025-04-09T06:53:23Z)
• Safe RLHF-V: Safe Reinforcement Learning from Multi-modal Human Feedback [34.01716144973483]
  Multimodal large language models (MLLMs) are essential for building general-purpose AI assistants.<n>How can we ensure safety alignment of MLLMs to prevent undesired behaviors?<n>In this work, we present the first exploration of the Safe RLHF-V -- the first multimodal safety alignment framework.
  arXiv  Detail & Related papers  (2025-03-22T07:40:20Z)
• Can Small Language Models Reliably Resist Jailbreak Attacks? A Comprehensive Evaluation [10.987263424166477]
  Small language models (SLMs) have emerged as promising alternatives to large language models (LLMs)<n>In this paper, we conduct the first large-scale empirical study of SLMs' vulnerabilities to jailbreak attacks.<n>We identify four key factors: model size, model architecture, training datasets and training techniques.
  arXiv  Detail & Related papers  (2025-03-09T08:47:16Z)
• SafeDialBench: A Fine-Grained Safety Benchmark for Large Language Models in Multi-Turn Dialogues with Diverse Jailbreak Attacks [90.41592442792181]
  We propose a fine-grained benchmark SafeDialBench for evaluating the safety of Large Language Models (LLMs)<n>Specifically, we design a two-tier hierarchical safety taxonomy that considers 6 safety dimensions and generates more than 4000 multi-turn dialogues in both Chinese and English under 22 dialogue scenarios.<n> Notably, we construct an innovative assessment framework of LLMs, measuring capabilities in detecting, and handling unsafe information and maintaining consistency when facing jailbreak attacks.
  arXiv  Detail & Related papers  (2025-02-16T12:08:08Z)
• Responsible AI in Construction Safety: Systematic Evaluation of Large Language Models and Prompt Engineering [9.559203170987598]
  Construction remains one of the most hazardous sectors. Recent advancements in AI, particularly Large Language Models (LLMs), offer promising opportunities for enhancing workplace safety. This study evaluates the performance of two widely used LLMs, GPT-3.5 and GPT-4o, across three standardized exams administered by the Board of Certified Safety Professionals (BCSP)
  arXiv  Detail & Related papers  (2024-11-13T04:06:09Z)
• What Makes and Breaks Safety Fine-tuning? A Mechanistic Study [64.9691741899956]
  Safety fine-tuning helps align Large Language Models (LLMs) with human preferences for their safe deployment. We design a synthetic data generation framework that captures salient aspects of an unsafe input. Using this, we investigate three well-known safety fine-tuning methods.
  arXiv  Detail & Related papers  (2024-07-14T16:12:57Z)
• Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training [67.30423823744506]
  We introduce a novel approach, Decoupled Refusal Training (DeRTa), designed to empower LLMs to refuse compliance to harmful prompts at any response position.<n>DeRTa incorporates two novel components: (1) Maximum Likelihood Estimation with Harmful Response Prefix, which trains models to recognize and avoid unsafe content by appending a segment of harmful response to the beginning of a safe response, and (2) Reinforced Transition Optimization (RTO), which equips models with the ability to transition from potential harm to safety refusal consistently throughout the harmful response sequence.
  arXiv  Detail & Related papers  (2024-07-12T09:36:33Z)

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.