Dual-Scale World Models for LLM Agents Towards Hard-Exploration Problems
This addresses the problem of inefficient exploration in LLM agents for interactive environments like text-based games, representing a strong incremental improvement.
The paper tackles the challenge of LLM-based agents struggling with hard-exploration tasks by introducing GLoW, a dual-scale world model approach that achieves state-of-the-art performance on the Jericho benchmark suite for text-based games, requiring 100-800x fewer environment interactions than RL-based methods.
LLM-based agents have seen promising advances, yet they are still limited in "hard-exploration" tasks requiring learning new knowledge through exploration. We present GLoW, a novel approach leveraging dual-scale world models, maintaining a trajectory frontier of high-value discoveries at the global scale, while learning from local trial-and-error in exploration through a Multi-path Advantage Reflection mechanism which infers advantage-based progress signals to guide exploration. To evaluate our framework for hard-exploration, we tackle the Jericho benchmark suite of text-based games, where GLoW achieves a new state-of-theart performance for LLM-based approaches. Compared to state-of-the-art RLbased methods, our approach achieves comparable performance while requiring 100-800x fewer environment interactions.