The Year in AI Papers
The year/Independent research

Paper 2511.04962

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

Zihao Yi, Qingxuan Jiang, Ruotian Ma, Xingyu Chen, Qu Yang, Mengru Wang, Fanghua Ye, Ying Shen, Zhaopeng Tu, Xiaolong Li, Linus

Published
Nov 2025
Research lab
Independent
Citations
7
GitHub
Not linked

01 In brief

Summary

This paper investigates the ability of large language models (LLMs) to role-play morally ambiguous or villainous characters, hypothesizing that safety alignment conflicts with authentic antagonistic portrayal.

The authors introduce the Moral RolePlay benchmark, a dataset with a four-level moral alignment scale (Moral Paragons, Flawed-but-Good, Egoists, Villains) and a balanced test set of 800 characters.

Evaluating state-of-the-art LLMs, they find a consistent, monotonic decline in role-playing fidelity as character morality decreases, with average scores dropping from 3.21 for paragons to 2.62 for villains.

The largest performance drop occurs between Level 2 and Level 3, indicating difficulty with self-serving personas.

Models struggle most with traits like 'Deceitful' and 'Manipulative', often substituting nuanced malevolence with superficial aggression.

General chatbot proficiency, as measured by Arena scores, is a poor predictor of villain role-playing ability, with highly safety-aligned models like Claude performing disproportionately poorly.

The study provides the first systematic evidence of this limitation, highlighting a tension between model safety and creative fidelity, and introduces the Villain RolePlay (VRP) leaderboard to rank models on this specific capability.

02 From the paper

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.