Socratic Planner: Self-QA-Based Zero-Shot Planning for Embodied Instruction Following

TL;DR

The Socratic Planner introduces a zero-shot, self-questioning approach using large language models to generate and adapt plans for embodied instruction following, outperforming existing methods on complex tasks and demonstrating real-world deployment.

Contribution

It presents a novel self-QA-based zero-shot planning method that generates and adjusts plans without training, enhancing embodied instruction following in complex environments.

Findings

Outperforms state-of-the-art models on ALFRED benchmark

Excels in long-horizon, complex inference tasks

Successfully deployed on a physical robot for real-world tasks

Abstract

Embodied Instruction Following (EIF) is the task of executing natural language instructions by navigating and interacting with objects in interactive environments. A key challenge in EIF is compositional task planning, typically addressed through supervised learning or few-shot in-context learning with labeled data. To this end, we introduce the Socratic Planner, a self-QA-based zero-shot planning method that infers an appropriate plan without any further training. The Socratic Planner first facilitates self-questioning and answering by the Large Language Model (LLM), which in turn helps generate a sequence of subgoals. While executing the subgoals, an embodied agent may encounter unexpected situations, such as unforeseen obstacles. The Socratic Planner then adjusts plans based on dense visual feedback through a visually-grounded re-planning mechanism. Experiments demonstrate the effectiveness of the Socratic Planner, outperforming current state-of-the-art planning models on the ALFRED benchmark across all metrics, particularly excelling in long-horizon tasks that demand complex inference. We further demonstrate its real-world applicability through deployment on a physical robot for long-horizon tasks.