Disentangling Intent from Role: Adversarial Self-Play for Persona-Invariant Safety Alignment

• URL: http://arxiv.org/abs/2605.01899v1
• Date: Sun, 03 May 2026 14:28:08 GMT
• Title: Disentangling Intent from Role: Adversarial Self-Play for Persona-Invariant Safety Alignment
• Abstract summary: Persona-Invariant Alignment (PIA) is an adversarial self-play framework that achieves co-evolution through Persona Lineage Evolution (PLE) on the attack side and Persona-Invariant Consistency Learning (PICL) on the defense side.<n> PICL is grounded in the structural separation hypothesis, using a unilateral KL-divergence constraint to enable the structural decoupling of safety decisions from persona context.<n> Experimental results demonstrate that PLE efficiently explores high-risk persona spaces by leveraging lineage-based credit propagation.
• Score: 13.780689172489934
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The growing capabilities of large language models (LLMs) have driven their widespread deployment across diverse domains, even in potentially high-risk scenarios. Despite advances in safety alignment techniques, current models remain vulnerable to emerging persona-based jailbreak attacks. Existing research on persona-based jailbreak has primarily focused on attack iterations, yet it lacks systemic and mechanistic constraints on the defense side. To address this challenge, we propose Persona-Invariant Alignment (PIA), an adversarial self-play framework that achieves co-evolution through Persona Lineage Evolution (PLE) on the attack side and Persona-Invariant Consistency Learning (PICL) on the defense side. Theoretically, PICL is grounded in the structural separation hypothesis, using a unilateral KL-divergence constraint to enable the structural decoupling of safety decisions from persona context, thereby maintaining safe behavior under persona-based jailbreak attacks. Experimental results demonstrate that PLE efficiently explores high-risk persona spaces by leveraging lineage-based credit propagation. Meanwhile, the PICL defense method significantly reduces the Attack Success Rate (ASR) while preserving the model's general capability, thereby validating the superiority and robustness of this alignment paradigm. Codes are available at https://github.com/JiajiaLi-1130/PIA.

Related papers

• Reasoning as an Attack Surface: Adaptive Evolutionary CoT Jailbreaks for LLMs [56.53007443197966]
  Large Reasoning Models (LRMs) have demonstrated remarkable capabilities in reasoning and generation tasks.<n>Their explicit chain-of-thought (CoT) mechanism introduces new security risks, making them particularly vulnerable to jailbreak attacks.<n>We propose an adaptive evolutionary CoT jailbreak framework, called AE-CoT, to overcome these limitations.
  arXiv  Detail & Related papers  (2026-05-23T10:11:32Z)
• SEMA: Simple yet Effective Learning for Multi-Turn Jailbreak Attacks [53.97948802255959]
  We propose a framework that trains a multi-turn attacker without relying on any existing strategies or external data.<n>Prefilling self-tuning enables usable rollouts by fine-tuning on non-refusal, well-structured, multi-turn adversarial prompts.<n>We anchor harmful intent in multi-turn jailbreaks via an intent-drift-aware reward that combines intent alignment, compliance risk, and level of detail.
  arXiv  Detail & Related papers  (2026-02-06T16:44:57Z)
• Stay in Character, Stay Safe: Dual-Cycle Adversarial Self-Evolution for Safety Role-Playing Agents [14.776514991797699]
  We propose a training-free Dual-Cycle Adversarial Self-Evolution framework.<n>A Persona-Targeted Attacker Cycle synthesizes progressively stronger jailbreak prompts.<n>A Role-Playing Defender Cycle distills observed failures into a hierarchical knowledge base.
  arXiv  Detail & Related papers  (2026-01-29T11:55:21Z)
• SafeBehavior: Simulating Human-Like Multistage Reasoning to Mitigate Jailbreak Attacks in Large Language Models [27.607151919652267]
  Large Language Models (LLMs) have achieved impressive performance across diverse natural language processing tasks.<n>But their growing power amplifies potential risks such as jailbreak attacks that circumvent built-in safety mechanisms.<n>We propose SafeBehavior, a novel hierarchical jailbreak defense mechanism that simulates the adaptive multistage reasoning process of humans.
  arXiv  Detail & Related papers  (2025-09-30T14:50:59Z)
• bi-GRPO: Bidirectional Optimization for Jailbreak Backdoor Injection on LLMs [33.470999703070866]
  Existing approaches to embedding jailbreak triggers suffer from limitations including poor generalization, compromised stealthiness, or reduced contextual usability.<n>We propose bi-GRPO, a novel RL-based framework tailored explicitly for jailbreak backdoor injection.
  arXiv  Detail & Related papers  (2025-09-24T05:56:41Z)
• Oyster-I: Beyond Refusal -- Constructive Safety Alignment for Responsible Language Models [93.5740266114488]
  Constructive Safety Alignment (CSA) protects against malicious misuse while actively guiding vulnerable users toward safe and helpful results.<n>Oy1 achieves state-of-the-art safety among open models while retaining high general capabilities.<n>We release Oy1, code, and the benchmark to support responsible, user-centered AI.
  arXiv  Detail & Related papers  (2025-09-02T03:04:27Z)
• DOPA: Stealthy and Generalizable Backdoor Attacks from a Single Client under Challenging Federated Constraints [2.139012072214621]
  Federated Learning (FL) is increasingly adopted for privacy-preserving collaborative training, but its decentralized nature makes it susceptible to backdoor attacks.<n>Existing attack methods, however, often rely on idealized assumptions and fail to remain effective under real-world constraints.<n>We propose DOPA, a novel framework that simulates heterogeneous local training dynamics and seeks consensus across divergent optimization trajectories to craft universally effective and stealthy backdoor triggers.
  arXiv  Detail & Related papers  (2025-08-20T08:39:12Z)
• Cannot See the Forest for the Trees: Invoking Heuristics and Biases to Elicit Irrational Choices of LLMs [83.11815479874447]
  We propose a novel jailbreak attack framework, inspired by cognitive decomposition and biases in human cognition.<n>We employ cognitive decomposition to reduce the complexity of malicious prompts and relevance bias to reorganize prompts.<n>We also introduce a ranking-based harmfulness evaluation metric that surpasses the traditional binary success-or-failure paradigm.
  arXiv  Detail & Related papers  (2025-05-03T05:28:11Z)
• Improving LLM Safety Alignment with Dual-Objective Optimization [81.98466438000086]
  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)
• Fundamental Limitations of Alignment in Large Language Models [16.393916864600193]
  An important aspect in developing language models that interact with humans is aligning their behavior to be useful and unharmful. This is usually achieved by tuning the model in a way that enhances desired behaviors and inhibits undesired ones. We propose a theoretical approach called Behavior Expectation Bounds (BEB) which allows us to formally investigate several inherent characteristics and limitations of alignment in large language models.
  arXiv  Detail & Related papers  (2023-04-19T17:50:09Z)
• A Self-supervised Approach for Adversarial Robustness [105.88250594033053]
  Adversarial examples can cause catastrophic mistakes in Deep Neural Network (DNNs) based vision systems. This paper proposes a self-supervised adversarial training mechanism in the input space. It provides significant robustness against the textbfunseen adversarial attacks.
  arXiv  Detail & Related papers  (2020-06-08T20:42:39Z)

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.