The Embodied World Model Based on LLM with Visual Information and Prediction-Oriented Prompts
This work addresses performance limitations in embodied AI systems for autonomous exploration in virtual environments, representing an incremental improvement over existing methods.
The study tackled the problem of underutilized visual data and insufficient world model functionality in LLM-based embodied AI, such as VOYAGER in Minecraft, by investigating the use of visual information and prediction-oriented prompts. The results showed that LLMs can extract necessary information from visual data, improving world model performance, and that devised prompts enhance this function.
In recent years, as machine learning, particularly for vision and language understanding, has been improved, research in embedded AI has also evolved. VOYAGER is a well-known LLM-based embodied AI that enables autonomous exploration in the Minecraft world, but it has issues such as underutilization of visual data and insufficient functionality as a world model. In this research, the possibility of utilizing visual data and the function of LLM as a world model were investigated with the aim of improving the performance of embodied AI. The experimental results revealed that LLM can extract necessary information from visual data, and the utilization of the information improves its performance as a world model. It was also suggested that devised prompts could bring out the LLM's function as a world model.