Preprint
GTBench: Uncovering the Strategic Reasoning Limitations of LLMs via Game-Theoretic Evaluations
arXiv.org
19 Feb 2024
Abstract
As Large Language Models (LLMs) are integrated into critical real-world applications, their strategic and logical reasoning abilities are increasingly crucial. This paper evaluates LLMs' reasoning abilities in competitive environments through game-theoretic tasks, e.g., board and card games that require pure logic and strategic reasoning to compete with opponents. We first propose GTBench, a language-driven environment composing 10 widely recognized tasks, across a comprehensive game taxonomy: complete versus incomplete information, dynamic versus static, and probabilistic versus deterministic scenarios. Then, we (1) Characterize the game-theoretic reasoning of LLMs; and (2) Perform LLM-vs.-LLM competitions as reasoning evaluation. We observe that (1) LLMs have distinct behaviors regarding various gaming scenarios; for example, LLMs fail in complete and deterministic games yet they are competitive in probabilistic gaming scenarios; (2) Most open-source LLMs, e.g., CodeLlama-34b-Instruct and Llama-2-70b-chat, are less competitive than commercial LLMs, e.g., GPT-4, in complex games, yet the recently released Llama-3-70b-Instruct makes up for this shortcoming. In addition, code-pretraining greatly benefits strategic reasoning, while advanced reasoning methods such as Chain-of-Thought (CoT) and Tree-of-Thought (ToT) do not always help. We further characterize the game-theoretic properties of LLMs, such as equilibrium and Pareto Efficiency in repeated games. Detailed error profiles are provided for a better understanding of LLMs' behavior. We hope our research provides standardized protocols and serves as a foundation to spur further explorations in the strategic reasoning of LLMs.
Metrics
15 Record Views
Details
- Title
- GTBench: Uncovering the Strategic Reasoning Limitations of LLMs via Game-Theoretic Evaluations
- Creators
- Jinhao Duan - Drexel UniversityRenming Zhang - Boston UniversityJames Diffenderfer - Landesamt für Landwirtschaft und nachhaltige LandentwicklungBhavya Kailkhura - Landesamt für Landwirtschaft und nachhaltige LandentwicklungLichao Sun - Lehigh UniversityElias Stengel-Eskin - University of North Carolina at Chapel HillMohit Bansal - University of North Carolina at Chapel HillTianlong Chen - University of North Carolina at Chapel HillKaidi Xu - Drexel University
- Publication Details
- arXiv.org
- Number of pages
- 29
- Resource Type
- Preprint
- Language
- English
- Academic Unit
- Computer Science
- Other Identifier
- 991021871355904721