# The Rise of Darkness: Safety-Utility Trade-Offs in Role-Playing Dialogue Agents

arXiv: https://arxiv.org/abs/2502.20757 · HTML: https://mlsanity.com/paper/2502.20757/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Placed in Agent threats, guardrails and governance, one of 16 directions in LLM agents and tool use.
- Builds on 30 of its 48 references that are in the map.
- First posted on arXiv 28 Feb 2025.

## Placed in

- LLM agents and tool use / [Agent threats, guardrails and governance](https://mlsanity.com/agents/agent-threats-guardrails-and-governance/)

## Builds on (most cited first, 30 of 30)

- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Generative Agents: Interactive Simulacra of Human Behavior](https://mlsanity.com/paper/2304.03442/) (2023)
- [QLoRA: Efficient Finetuning of Quantized LLMs](https://mlsanity.com/paper/2305.14314/) (2023)
- [Jailbroken: How Does LLM Safety Training Fail?](https://mlsanity.com/paper/2307.02483/) (2023)
- [RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models](https://mlsanity.com/paper/2009.11462/) (2020)
- [Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!](https://mlsanity.com/paper/2310.03693/) (2023)
- [SimPO: Simple Preference Optimization with a Reference-Free Reward](https://mlsanity.com/paper/2405.14734/) (2024)
- [A Survey on In-context Learning](https://mlsanity.com/paper/2301.00234/) (2022)
- [ORPO: Monolithic Preference Optimization without Reference Model](https://mlsanity.com/paper/2403.07691/) (2024)
- [Toxicity in ChatGPT: Analyzing Persona-assigned Language Models](https://mlsanity.com/paper/2304.05335/) (2023)
- [Character-LLM: A Trainable Agent for Role-Playing](https://mlsanity.com/paper/2310.10158/) (2023)
- [InCharacter: Evaluating Personality Fidelity in Role-Playing Agents through Psychological Interviews](https://mlsanity.com/paper/2310.17976/) (2023)
- [Rewards-in-Context: Multi-objective Alignment of Foundation Models with Dynamic Preference Adjustment](https://mlsanity.com/paper/2402.10207/) (2024)
- [Beyond One-Preference-Fits-All Alignment: Multi-Objective Direct Preference Optimization](https://mlsanity.com/paper/2310.03708/) (2023)
- [Language Models are Homer Simpson! Safety Re-Alignment of Fine-tuned Language Models through Task Arithmetic](https://mlsanity.com/paper/2402.11746/) (2024)
- [Large Language Models are Superpositions of All Characters: Attaining Arbitrary Role-play via Self-Alignment](https://mlsanity.com/paper/2401.12474/) (2024)
- [Multi-Party Chat: Conversational Agents in Group Settings with Humans and Models](https://mlsanity.com/paper/2304.13835/) (2023)
- [Neeko: Leveraging Dynamic LoRA for Efficient Multi-Character Role-Playing Agent](https://mlsanity.com/paper/2402.13717/) (2024)
- [SocialBench: Sociality Evaluation of Role-Playing Conversational Agents](https://mlsanity.com/paper/2403.13679/) (2024)
- [The Oscars of AI Theater: A Survey on Role-Playing with Language Models](https://mlsanity.com/paper/2407.11484/) (2024)
- [Harnessing the Power of Large Language Models for Empathetic Response Generation: Empirical Investigations and Improvements](https://mlsanity.com/paper/2310.05140/) (2023)
- [From Role-Play to Drama-Interaction: An LLM Solution](https://mlsanity.com/paper/2405.14231/) (2024)
- [Emulated Disalignment: Safety Alignment for Large Language Models May Backfire!](https://mlsanity.com/paper/2402.12343/) (2024)
- [IBSEN: Director-Actor Agent Collaboration for Controllable and Interactive Drama Script Generation](https://mlsanity.com/paper/2407.01093/) (2024)
- [Exploring Safety-Utility Trade-Offs in Personalized Language Models](https://mlsanity.com/paper/2406.11107/) (2024)
- [Enhancing Personalized Dialogue Generation with Contrastive Latent Variables: Combining Sparse and Dense Persona](https://mlsanity.com/paper/2305.11482/) (2023)
- [Identity-Driven Hierarchical Role-Playing Agents](https://mlsanity.com/paper/2407.19412/) (2024)
- [A Multi-Task Role-Playing Agent Capable of Imitating Character Linguistic Styles](https://mlsanity.com/paper/2411.02457/) (2024)
- [ERABAL: Enhancing Role-Playing Agents through Boundary-Aware Learning](https://mlsanity.com/paper/2409.14710/) (2024)
- [On the Hallucination in Simultaneous Machine Translation](https://mlsanity.com/paper/2406.07239/) (2024)

## Built on by (most cited first, 7 of 7)

- [Generator-Assistant Stepwise Rollback Framework for Large Language Model Agent](https://mlsanity.com/paper/2503.02519/) (2025)
- [ORPP: Self-Optimizing Role-playing Prompts to Enhance Language Model Capabilities](https://mlsanity.com/paper/2506.02480/) (2025)
- [Too Nice to Tell the Truth: Quantifying Agreeableness-Driven Sycophancy in Role-Playing Language Models](https://mlsanity.com/paper/2604.10733/) (2026)
- [MASTER: Enhancing Large Language Model via Multi-Agent Simulated Teaching](https://mlsanity.com/paper/2506.02689/) (2025)
- [Persona Non Grata: Single-Method Safety Evaluation Is Incomplete for Persona-Imbued LLMs](https://mlsanity.com/paper/2604.11120/) (2026)
- [Do LLM Agents Mirror Socio-Cognitive Effects in Power-Asymmetric Conversations?](https://mlsanity.com/paper/2605.17694/) (2026)
- [User-Aware Active Knowledge Acquisition for Emotional Support Dialogue](https://mlsanity.com/paper/2605.29715/) (2026)

## Abstract (arXiv)

Large Language Models (LLMs) have made remarkable advances in role-playing dialogue agents, demonstrating their utility in character simulations. However, it remains challenging for these agents to balance character portrayal utility with content safety because this essential character simulation often comes with the risk of generating unsafe content. To address this issue, we first conduct a systematic exploration of the safety-utility trade-off across multiple LLMs. Our analysis reveals that risk scenarios created by villain characters and user queries (referred to as risk coupling) contribute to this trade-off. Building on this, we propose a novel Adaptive Dynamic Multi-Preference (ADMP) method, which dynamically adjusts safety-utility preferences based on the degree of risk coupling and guides the model to generate responses biased toward utility or safety. We further introduce Coupling Margin Sampling (CMS) into coupling detection to enhance the model's ability to handle high-risk scenarios. Experimental results demonstrate that our approach improves safety metrics while maintaining utility.