This paper explores the potential of Large Language Models (LLMs) as neurosymbolic reasoners, focusing on text-based games that involve symbolic tasks such as math, map reading, sorting, and applying common sense. The authors propose an LLM agent designed to tackle these symbolic challenges and achieve in-game objectives. The agent is initialized with its role and receives observations, valid actions, and symbolic modules from the game environment. Using these inputs, the LLM agent selects actions to interact with the game environment and symbolic modules, enhancing its performance in text-based games. Experimental results demonstrate that the LLM agent outperforms strong baselines, achieving an average performance of 88% across all tasks. The paper highlights the effectiveness of LLMs in symbolic reasoning and suggests that they can be considered neurosymbolic reasoners, with significant potential for real-world applications.This paper explores the potential of Large Language Models (LLMs) as neurosymbolic reasoners, focusing on text-based games that involve symbolic tasks such as math, map reading, sorting, and applying common sense. The authors propose an LLM agent designed to tackle these symbolic challenges and achieve in-game objectives. The agent is initialized with its role and receives observations, valid actions, and symbolic modules from the game environment. Using these inputs, the LLM agent selects actions to interact with the game environment and symbolic modules, enhancing its performance in text-based games. Experimental results demonstrate that the LLM agent outperforms strong baselines, achieving an average performance of 88% across all tasks. The paper highlights the effectiveness of LLMs in symbolic reasoning and suggests that they can be considered neurosymbolic reasoners, with significant potential for real-world applications.