Self-Corrective Task Planning by Inverse Prompting with Large Language Models

2025-03-10

AI Summary

In long-horizon robotic task planning, large language models (LLMs) often generate action sequences that are logically consistent yet factually incorrect—i.e., “hallucinations”—while existing self-correction methods fail due to insufficient reasoning depth. To address this, we propose a reverse-prompting–based self-correction framework. Its core innovation is a novel reverse-action verification mechanism: by generating and simulating backward actions, the framework explicitly checks whether the original plan can restore the initial state, thereby jointly enforcing logical consistency and enabling interpretable feedback. The method integrates reverse prompting, stepwise chain-of-thought reasoning, and action-state reversibility modeling, supporting LLM-driven iterative refinement. Evaluated on standard benchmarks, our approach achieves a 16.3% average success rate improvement over state-of-the-art LLM-based task planners and self-correction methods, demonstrating superior robustness and interpretability.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Planning with Language Models
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

In robot task planning, large language models (LLMs) have shown significant promise in generating complex and long-horizon action sequences. However, it is observed that LLMs often produce responses that sound plausible but are not accurate. To address these problems, existing methods typically employ predefined error sets or external knowledge sources, requiring human efforts and computation resources. Recently, self-correction approaches have emerged, where LLM generates and refines plans, identifying errors by itself. Despite their effectiveness, they are more prone to failures in correction due to insufficient reasoning. In this paper, we introduce InversePrompt, a novel self-corrective task planning approach that leverages inverse prompting to enhance interpretability. Our method incorporates reasoning steps to provide clear, interpretable feedback. It generates inverse actions corresponding to the initially generated actions and verifies whether these inverse actions can restore the system to its original state, explicitly validating the logical coherence of the generated plans. The results on benchmark datasets show an average 16.3% higher success rate over existing LLM-based task planning methods. Our approach offers clearer justifications for feedback in real-world environments, resulting in more successful task completion than existing self-correction approaches across various scenarios.

Problem

• Improves robot task planning accuracy using LLMs.
• Enhances interpretability with inverse prompting and reasoning steps.
• Increases success rate by validating logical coherence of plans.

Innovation

• InversePrompt enhances interpretability in task planning.
• Generates inverse actions to validate logical coherence.
• Improves success rate by 16.3% over existing methods.