Too Good to be Bad: On the Failure of LLMs to Role-Play Villains

• URL: http://arxiv.org/abs/2511.04962v2
• Date: Thu, 13 Nov 2025 01:43:18 GMT
• Title: Too Good to be Bad: On the Failure of LLMs to Role-Play Villains
• Authors: Zihao Yi, Qingxuan Jiang, Ruotian Ma, Xingyu Chen, Qu Yang, Mengru Wang, Fanghua Ye, Ying Shen, Zhaopeng Tu, Xiaolong Li, Linus,
• Abstract summary: Large Language Models (LLMs) are increasingly tasked with creative generation, including the simulation of fictional characters.<n>We hypothesize that the safety alignment of modern LLMs creates a fundamental conflict with the task of authentically role-playing morally ambiguous or villainous characters.<n>We introduce the Moral RolePlay benchmark, a new dataset featuring a four-level moral alignment scale and a balanced test set for rigorous evaluation.<n>Our large-scale evaluation reveals a consistent, monotonic decline in role-playing fidelity as character morality decreases.
• Score: 69.0500092126915
• License: http://creativecommons.org/publicdomain/zero/1.0/
• Abstract: Large Language Models (LLMs) are increasingly tasked with creative generation, including the simulation of fictional characters. However, their ability to portray non-prosocial, antagonistic personas remains largely unexamined. We hypothesize that the safety alignment of modern LLMs creates a fundamental conflict with the task of authentically role-playing morally ambiguous or villainous characters. To investigate this, we introduce the Moral RolePlay benchmark, a new dataset featuring a four-level moral alignment scale and a balanced test set for rigorous evaluation. We task state-of-the-art LLMs with role-playing characters from moral paragons to pure villains. Our large-scale evaluation reveals a consistent, monotonic decline in role-playing fidelity as character morality decreases. We find that models struggle most with traits directly antithetical to safety principles, such as ``Deceitful'' and ``Manipulative'', often substituting nuanced malevolence with superficial aggression. Furthermore, we demonstrate that general chatbot proficiency is a poor predictor of villain role-playing ability, with highly safety-aligned models performing particularly poorly. Our work provides the first systematic evidence of this critical limitation, highlighting a key tension between model safety and creative fidelity. Our benchmark and findings pave the way for developing more nuanced, context-aware alignment methods.

Related papers

• Are Language Models Sensitive to Morally Irrelevant Distractors? [47.92026843851412]
  We show that moral distractors can shift the moral judgements of large language models by over 30% even in low-ambiguity scenarios.<n>This research challenges theories that assume the stability of human moral judgements.
  arXiv  Detail & Related papers  (2026-02-10T05:18:05Z)
• Do VLMs Have a Moral Backbone? A Study on the Fragile Morality of Vision-Language Models [41.633874062439254]
  It remains unclear whether Vision-Language Models (VLMs) are stable in realistic settings.<n>We probe VLMs with a diverse set of model-agnostic multimodal perturbations and find that their moral stances are highly fragile.<n>We show that lightweight inference-time interventions can partially restore moral stability.
  arXiv  Detail & Related papers  (2026-01-23T06:00:09Z)
• The Straight and Narrow: Do LLMs Possess an Internal Moral Path? [25.256151938852728]
  Current alignment techniques often act as superficial guardrails, leaving the intrinsic moral representations of Large Language Models largely untouched.<n>We bridge this gap by leveraging Moral Foundations Theory (MFT) to map and manipulate the fine-grained moral landscape of LLMs.<n>We propose Adaptive Moral Fusion (AMF), a dynamic inference-time intervention that synergizes probe detection with vector injection to tackle the safety-helpfulness trade-off.
  arXiv  Detail & Related papers  (2026-01-15T11:42:00Z)
• Evaluating from Benign to Dynamic Adversarial: A Squid Game for Large Language Models [57.33350664910483]
  We introduce Squid Game, a dynamic and adversarial evaluation environment with resource-constrained and asymmetric information settings.<n>We evaluate over 50 LLMs on Squid Game, presenting the largest behavioral evaluation study of general LLMs on dynamic adversarial scenarios.
  arXiv  Detail & Related papers  (2025-11-12T06:06:29Z)
• Moral Susceptibility and Robustness under Persona Role-Play in Large Language Models [0.0]
  We introduce a benchmark that quantifies two properties: moral susceptibility and moral robustness.<n>For moral robustness, model family accounts for most of the variance, while model size shows no systematic effect.<n>Moral susceptibility exhibits a mild family effect but a clear within-family size effect, with larger variants being more susceptible.
  arXiv  Detail & Related papers  (2025-11-11T18:47:44Z)
• Beyond One World: Benchmarking Super Heros in Role-Playing Across Multiversal Contexts [2.2816872489992135]
  We introduce Beyond One World, a benchmark for character-grounded roleplay spanning 30 iconic heroes and 90 canon-specific versions.<n>We score responses for canonical accuracy and reasoning fidelity.<n>We propose Think-Act Matching, a metric that quantifies alignment between reasons and actions.
  arXiv  Detail & Related papers  (2025-10-16T06:39:27Z)
• MORABLES: A Benchmark for Assessing Abstract Moral Reasoning in LLMs with Fables [50.29407048003165]
  We present MORABLES, a human-verified benchmark built from fables and short stories drawn from historical literature.<n>The main task is structured as multiple-choice questions targeting moral inference, with carefully crafted distractors that challenge models to go beyond shallow, extractive question answering.<n>Our findings show that, while larger models outperform smaller ones, they remain susceptible to adversarial manipulation and often rely on superficial patterns rather than true moral reasoning.
  arXiv  Detail & Related papers  (2025-09-15T19:06:10Z)
• Beyond Ethical Alignment: Evaluating LLMs as Artificial Moral Assistants [0.36326779753373206]
  The recent rise in popularity of large language models (LLMs) has prompted considerable concerns about their moral capabilities.<n>This paper examines their capacity to function as Artificial Moral Assistants (AMAs)<n>We argue that qualifying as an AMA requires more than what state-of-the-art alignment techniques aim to achieve.
  arXiv  Detail & Related papers  (2025-08-18T09:28:55Z)
• When Ethics and Payoffs Diverge: LLM Agents in Morally Charged Social Dilemmas [68.79830818369683]
  Large language models (LLMs) have enabled their use in complex agentic roles, involving decision-making with humans or other agents.<n>Recent advances in large language models (LLMs) have enabled their use in complex agentic roles, involving decision-making with humans or other agents.<n>There is limited understanding of how they act when moral imperatives directly conflict with rewards or incentives.<n>We introduce Moral Behavior in Social Dilemma Simulation (MoralSim) and evaluate how LLMs behave in the prisoner's dilemma and public goods game with morally charged contexts.
  arXiv  Detail & Related papers  (2025-05-25T16:19:24Z)
• M$^3$oralBench: A MultiModal Moral Benchmark for LVLMs [66.78407469042642]
  We introduce M$3$oralBench, the first MultiModal Moral Benchmark for LVLMs.<n>M$3$oralBench expands the everyday moral scenarios in Moral Foundations Vignettes (MFVs) and employs the text-to-image diffusion model, SD3.0, to create corresponding scenario images.<n>It conducts moral evaluation across six moral foundations of Moral Foundations Theory (MFT) and encompasses tasks in moral judgement, moral classification, and moral response.
  arXiv  Detail & Related papers  (2024-12-30T05:18:55Z)
• When to Make Exceptions: Exploring Language Models as Accounts of Human Moral Judgment [96.77970239683475]
  AI systems need to be able to understand, interpret and predict human moral judgments and decisions. A central challenge for AI safety is capturing the flexibility of the human moral mind. We present a novel challenge set consisting of rule-breaking question answering.
  arXiv  Detail & Related papers  (2022-10-04T09:04:27Z)
• Moral Mimicry: Large Language Models Produce Moral Rationalizations Tailored to Political Identity [0.0]
  This study investigates whether Large Language Models reproduce the moral biases associated with political groups in the United States. Using tools from Moral Foundations Theory, it is shown that these LLMs are indeed moral mimics.
  arXiv  Detail & Related papers  (2022-09-24T23:55:53Z)

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.