Evaluating World Models with LLM for Decision Making
This work provides a benchmark for evaluating LLM-based world models in decision-making, which is incremental as it builds on existing environments and tasks to assess performance rather than introducing new methods.
The paper tackles the problem of evaluating world models based on Large Language Models (LLMs) for decision-making by proposing a comprehensive evaluation framework across 31 diverse environments and three tasks: policy verification, action proposal, and policy planning. The results show that GPT-4o significantly outperforms GPT-4o-mini, especially in tasks requiring domain knowledge, but performance decreases for long-term decision-making and combining functionalities introduces instability.
World model emerges as a key module in decision making, where MuZero and Dreamer achieve remarkable successes in complex tasks. Recent work leverages Large Language Models (LLMs) as general world simulators to simulate the dynamics of the world due to their generalizability. LLMs also serve as the world model for deliberative reasoning in Reasoning via Planning (RAP) and Tree of Thought (ToT). However, the world models are either evaluated as a general world simulator, or as a functional module of the agent, i.e., predicting the transitions to assist the planning. In this work, we propose a comprehensive evaluation of the world models with LLMs from the decision making perspective. Specifically, we leverage the 31 diverse environments from (Wang et al., 2023;2024) and curate the rule-based policy of each environment for the diverse evaluation. Then, we design three main tasks, i.e., policy verification, action proposal, and policy planning, where the world models can be used for decision making solely. Finally, we conduct the comprehensive evaluation of the advanced LLMs, i.e., GPT-4o and GPT-4o-mini, on the environments for the three main tasks under various settings. The key observations include: i) GPT-4o significantly outperforms GPT-4o-mini on the three main tasks, especially for the tasks which require the domain knowledge, ii) the performance of the world model with LLM will be decreased for long-term decision-making tasks, and iii) the combination of different functionalities of the world model will brings additional unstabilities of the performance.