=Overview

Abstract: Effective planning in the real world requires not only world knowledge, but the ability to leverage that knowledge to build the right representation of the task at hand. Decades of hierarchical planning techniques have used domain-specific temporal action abstractions to support efficient and accurate planning, almost always relying on human priors and domain knowledge to decompose hard tasks into smaller subproblems appropriate for a goal or set of goals. This paper describes Ada (Action Domain Acquisition), a framework for automatically constructing task-specific planning representations using task-general background knowledge from language models (LMs). Starting with a general-purpose hierarchical planner and a low-level goal-conditioned policy, Ada interactively learns a library of planner-compatible high-level action abstractions and low-level controllers adapted to a particular domain of planning tasks. On two language-guided interactive planning benchmarks (Mini Minecraft and ALFRED Household Tasks), Ada strongly outperforms other approaches that use LMs for sequential decision-making, offering more accurate plans and better generalization to complex tasks.

=Training

The training phase involves constructing task-specific planning representations by leveraging task-general background knowledge from language models. The framework learns high-level action abstractions and low-level controllers adapted to the domain of interest, enabling efficient and accurate planning.

=Inference

During inference, the system applies the learned representations and policies to solve complex planning tasks. It generalizes effectively to unseen scenarios, creating plans that adapt to task-specific requirements in real time.

=Learned Operator in the ALFRED Environment

=Learned Operator in the Mini-Mine Environment

=Result