Saved in:
Bibliographic Details
Main Authors: Alyahya, Hisham A., Khan, Haidar, Alnumay, Yazeed, Bari, M Saiful, Yener, Bülent
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2503.10673
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908323064840192
author Alyahya, Hisham A.
Khan, Haidar
Alnumay, Yazeed
Bari, M Saiful
Yener, Bülent
author_facet Alyahya, Hisham A.
Khan, Haidar
Alnumay, Yazeed
Bari, M Saiful
Yener, Bülent
contents We introduce ZeroSumEval, a dynamic, competition-based, and evolving evaluation framework for Large Language Models (LLMs) that leverages competitive games. ZeroSumEval encompasses a diverse suite of games, including security challenges (Capture the Flag), classic board games (chess), and knowledge tests (MathQuiz). These games are designed to evaluate a range of capabilities such as strategic reasoning, planning, knowledge application, safety, and adaptability. Building upon recent studies that highlight the effectiveness of game-based evaluations for LLMs, ZeroSumEval enhances these approaches by providing a standardized and extensible framework for easily implementing games and leverages DSPy to provide a better abstraction for LLM player strategies.
format Preprint
id arxiv_https___arxiv_org_abs_2503_10673
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ZeroSumEval: An Extensible Framework For Scaling LLM Evaluation with Inter-Model Competition
Alyahya, Hisham A.
Khan, Haidar
Alnumay, Yazeed
Bari, M Saiful
Yener, Bülent
Computation and Language
Artificial Intelligence
We introduce ZeroSumEval, a dynamic, competition-based, and evolving evaluation framework for Large Language Models (LLMs) that leverages competitive games. ZeroSumEval encompasses a diverse suite of games, including security challenges (Capture the Flag), classic board games (chess), and knowledge tests (MathQuiz). These games are designed to evaluate a range of capabilities such as strategic reasoning, planning, knowledge application, safety, and adaptability. Building upon recent studies that highlight the effectiveness of game-based evaluations for LLMs, ZeroSumEval enhances these approaches by providing a standardized and extensible framework for easily implementing games and leverages DSPy to provide a better abstraction for LLM player strategies.
title ZeroSumEval: An Extensible Framework For Scaling LLM Evaluation with Inter-Model Competition
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2503.10673