Multicultural Spyfall: Assessing LLMs through Dynamic Multilingual Social Deduction Game

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Wibowo, Haryo Akbarianto, Elsetohy, Alaa, Cui, Qinrong, Aji, Alham Fikri
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866918288236216320
author Wibowo, Haryo Akbarianto
Elsetohy, Alaa
Cui, Qinrong
Aji, Alham Fikri
author_facet Wibowo, Haryo Akbarianto
Elsetohy, Alaa
Cui, Qinrong
Aji, Alham Fikri
contents The rapid advancement of Large Language Models (LLMs) has necessitated more robust evaluation methods that go beyond static benchmarks, which are increasingly prone to data saturation and leakage. In this paper, we propose a dynamic benchmarking framework for evaluating multilingual and multicultural capabilities through the social deduction game Spyfall. In our setup, models must engage in strategic dialogue to either identify a secret agent or avoid detection, utilizing culturally relevant locations or local foods. Our results show that our game-based rankings align closely with the Chatbot Arena. However, we find a significant performance gap in non-English contexts: models are generally less proficient when handling locally specific entities and often struggle with rule-following or strategic integrity in non-English languages. We demonstrate that this game-based approach provides a scalable, leakage-resistant, and culturally nuanced alternative to traditional NLP benchmarks. The game history can be accessed here https://huggingface.co/datasets/haryoaw/cultural-spyfall.
format Preprint
id arxiv_https___arxiv_org_abs_2601_09017
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Multicultural Spyfall: Assessing LLMs through Dynamic Multilingual Social Deduction Game
Wibowo, Haryo Akbarianto
Elsetohy, Alaa
Cui, Qinrong
Aji, Alham Fikri
Computation and Language
68T50
The rapid advancement of Large Language Models (LLMs) has necessitated more robust evaluation methods that go beyond static benchmarks, which are increasingly prone to data saturation and leakage. In this paper, we propose a dynamic benchmarking framework for evaluating multilingual and multicultural capabilities through the social deduction game Spyfall. In our setup, models must engage in strategic dialogue to either identify a secret agent or avoid detection, utilizing culturally relevant locations or local foods. Our results show that our game-based rankings align closely with the Chatbot Arena. However, we find a significant performance gap in non-English contexts: models are generally less proficient when handling locally specific entities and often struggle with rule-following or strategic integrity in non-English languages. We demonstrate that this game-based approach provides a scalable, leakage-resistant, and culturally nuanced alternative to traditional NLP benchmarks. The game history can be accessed here https://huggingface.co/datasets/haryoaw/cultural-spyfall.
title Multicultural Spyfall: Assessing LLMs through Dynamic Multilingual Social Deduction Game
topic Computation and Language
68T50
url https://arxiv.org/abs/2601.09017