mlsanityData to 7 Oct 2026

Simple Role Assignment is Extraordinarily Effective for Safety Alignment

arXiv 2602.00061 ↗First posted 20 Jan 2026cs.CY cs.AIcited by 1

On the map: LLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks

also in 1 more
Abstract (arXiv)

Principle-based alignment often lacks context sensitivity and completeness. Grounded in Theory of Mind, we propose role conditioning as a compact alternative: social roles (e.g., mother, judge) implicitly encode both values and the cognitive schemas required to apply them. We introduce a training-free pipeline featuring a role-conditioned generator and iterative role-based critics for refinement. Across five model families, our approach consistently outperforms principle-based, Chain-of-Thought (CoT) and other baselines across benchmarks. Notably, it reduces unsafe outputs on the WildJailbreak benchmark from 81.4% to 3.6% with DeepSeek-V3. Not only for common safety benchmarks, it consistently applies for agentic safety tasks. These results establish role assignment as a powerful, interpretable paradigm for AI alignment and LLM-as-a-Judge construction.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in Safety alignment and jailbreak attacks, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 30 of its 49 references that are in the map. First posted on arXiv 20 Jan 2026.

Builds on

30 of 49 references.

Closest ancestors · 1

Same direction, newest first

  1. 2023Universal and Transferable Adversarial Attacks on Aligned Language Models27 Jul

Other references · 29

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  3. 2017Deep reinforcement learning from human preferencescited by 3,846
  4. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,655
  5. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 2,950
  6. 2024SimPO: Simple Preference Optimization with a Reference-Free Rewardcited by 951
  7. 2024Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluatorscited by 865
  8. 2023Character-LLM: A Trainable Agent for Role-Playingcited by 335
  9. 2024SALAD-Bench: A Hierarchical and Comprehensive Safety Benchmark for Large Language Modelscited by 225
  10. 2024ShieldGemma: Generative AI Content Moderation Based on Gemmacited by 224
  11. 2023Better Zero-Shot Reasoning with Role-Play Promptingcited by 220
  12. 2023Red-Teaming Large Language Models using Chain of Utterances for Safety-Alignmentcited by 213
17 more
  1. 2024Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judgecited by 202
  2. 2023RAIN: Your Language Models Can Align Themselves without Finetuningcited by 149
  3. 2024Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templatescited by 94
  4. 2024Self-Alignment for Factuality: Mitigating Hallucinations in LLMs via Self-Evaluationcited by 79
  5. 2024Open-LLM-Leaderboard: From Multi-choice to Open-style Questions for LLMs Evaluation, Benchmark, and Arenacited by 76
  6. 2024Roleplay-doh: Enabling Domain-Experts to Create LLM-simulated Patients via Eliciting and Adhering to Principlescited by 72
  7. 2024LLM Discussion: Enhancing the Creativity of Large Language Models via Discussion Framework and Role-Playcited by 65
  8. 2024Self-Alignment of Large Language Models via Monopolylogue-based Social Scene Simulationcited by 46
  9. 2025MCTS-Judge: Test-Time Scaling in LLM-as-a-Judge for Code Correctness Evaluationcited by 32
  10. 2025CodeJudgeBench: Benchmarking LLM-as-a-Judge for Coding Taskscited by 32
  11. 2024CodeJudge-Eval: Can Large Language Models be Good Judges in Code Understanding?cited by 29
  12. 2024Critic-CoT: Boosting the reasoning abilities of large language model via Chain-of-thoughts Criticcited by 27
  13. 2025Safer or Luckier? LLMs as Safety Evaluators Are Not Robust to Artifactscited by 25
  14. 2024Aligning Large Language Models by On-Policy Self-Judgmentcited by 20
  15. 2025Potential and Perils of Large Language Models as Judges of Unstructured Textual Datacited by 11
  16. 2025Improving LLM Reasoning through Interpretable Role-Playing Steeringcited by 11
  17. 2025On-the-fly Preference Alignment via Principle-Guided Decodingcited by 10

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Credibility Governance: A Social Mechanism for Collective Self-Correction under Weak Truth Signalscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025AgentAlign: Navigating Safety Alignment in the Shift from Informative to Agentic Large Language Models
  2. 2026Beyond Refusal Patterns: Safe-Role Internalization for Robust and Generalizable LLM Safety Alignment
  3. 2026Reducing the Safety Tax in LLM Safety Alignment with On-Policy Self-Distillation
  4. 2024Combining Theory of Mind and Kindness for Self-Supervised Human-AI Alignment
  5. 2026Rationalize: Shared Semantic Reasoning for Human-AI Alignment
  6. 2026DeShortcut-Align: Decoupling Spurious Shortcuts for Robust Safety Alignment in Large Reasoning Models
  7. 2025SafeWork-R1: Coevolving Safety and Intelligence under the AI-45circ Law
  8. 2025Self-Guided Adaptive Safety Alignment: Synthesizing and Internalizing Guidelines in Reasoning Models