Learning with Language-Guided State Abstractions

arXiv:2402.18759v1 Announce Type: cross
Abstract: We describe a framework for using natural language to design state abstractions for imitation learning. Generalizable policy learning in high-dimensional observation spaces is facilitated by well-designed state representations, which can surface important features of an environment and hide irrelevant ones. These state representations are typically manually specified, or derived from other labor-intensive labeling procedures. Our method, LGA (language-guided abstraction), uses a combination of natural language supervision and background knowledge from language models (LMs) to automatically …

abstract abstractions arxiv cs.ai cs.lg cs.ro design environment features framework hide imitation learning language natural natural language observation policy spaces state surface type