DUPLEX: Agentic Dual-System Planning via LLM-Driven Information Extraction

TL;DR

DUPLEX introduces a dual-system architecture that combines a lightweight LLM for semantic information extraction with a symbolic planner for reliable long-horizon task planning, significantly improving success rates in various domains.

Contribution

This work presents a neuro-symbolic framework that confines LLMs to information extraction, enhancing planning reliability by integrating classical symbolic planning with LLM-driven diagnostics.

Findings

Outperforms existing LLM-based planning methods in success rate and reliability.

Effectively handles complex and underspecified scenarios through iterative reflection.

Demonstrates robustness across 12 classical and household planning domains.

Abstract

While Large Language Models (LLMs) provide semantic flexibility for robotic task planning, their susceptibility to hallucination and logical inconsistency limits their reliability in long-horizon domains. To bridge the gap between unstructured environments and rigorous plan synthesis, we propose DUPLEX, an agentic dual-system neuro-symbolic architecture that strictly confines the LLM to schema-guided information extraction rather than end-to-end planning or code generation. In our framework, a feed-forward Fast System utilizes a lightweight LLM to extract entities, relations etc. from natural language, deterministically mapping them into a Planning Domain Definition Language (PDDL) problem file for a classical symbolic planner. To resolve complex or underspecified scenarios, a Slow System is activated exclusively upon planning failure, leveraging solver diagnostics to drive a high-capacity LLM in iterative reflection and repair. Extensive evaluations across 12 classical and household planning domains demonstrate that DUPLEX significantly outperforms existing end-to-end and hybrid LLM baselines in both success rate and reliability. These results confirm that The key is not to make the LLM plan better, but to restrict the LLM to the part it is good at - structured semantic grounding - and leave logical plan synthesis to a symbolic planner.