NeSyPr: Neurosymbolic Proceduralization For Efficient Embodied Reasoning

2025-10-22

AI Summary

In dynamic environments, large language models (LMs) face significant challenges in online invoking large-scale inference engines or symbolic planners due to latency, connectivity, and resource constraints—severely limiting embodied task performance. To address this, we propose NeSyPr, a neuro-symbolic programming framework that compiles declarative planning knowledge into programmable rules directly integrable into LMs. NeSyPr abstracts and generalizes multi-step symbolic reasoning into single-step neural inference via neuro-symbolic programming, eliminating runtime dependence on external symbolic systems. Its core contribution lies in embedding declarative planning knowledge into LM inference, enabling generation of composable, structured, programmatic reasoning paths. Evaluated on PDDLGym, VirtualHome, and ALFWorld benchmarks, NeSyPr with only medium- or small-scale LMs surpasses both large reasoning models and pure symbolic planners, achieving substantial improvements in both inference efficiency and task success rates.

Technology Category

• Machine Learning: Neuro-Symbolic Learning
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

We address the challenge of adopting language models (LMs) for embodied tasks in dynamic environments, where online access to large-scale inference engines or symbolic planners is constrained due to latency, connectivity, and resource limitations. To this end, we present NeSyPr, a novel embodied reasoning framework that compiles knowledge via neurosymbolic proceduralization, thereby equipping LM-based agents with structured, adaptive, and timely reasoning capabilities. In NeSyPr, task-specific plans are first explicitly generated by a symbolic tool leveraging its declarative knowledge. These plans are then transformed into composable procedural representations that encode the plans' implicit production rules, enabling the resulting composed procedures to be seamlessly integrated into the LM's inference process. This neurosymbolic proceduralization abstracts and generalizes multi-step symbolic structured path-finding and reasoning into single-step LM inference, akin to human knowledge compilation. It supports efficient test-time inference without relying on external symbolic guidance, making it well suited for deployment in latency-sensitive and resource-constrained physical systems. We evaluate NeSyPr on the embodied benchmarks PDDLGym, VirtualHome, and ALFWorld, demonstrating its efficient reasoning capabilities over large-scale reasoning models and a symbolic planner, while using more compact LMs.

Problem

• Enabling efficient embodied reasoning with limited computational resources
• Compiling symbolic plans into procedural LM-integrable representations
• Achieving real-time inference without external symbolic guidance

Innovation

• Compiles symbolic plans into procedural representations
• Integrates composed procedures into LM inference
• Enables single-step LM inference without external guidance