NeSyPr: Neurosymbolic Proceduralization For Efficient Embodied Reasoning

TL;DR

NeSyPr introduces neurosymbolic proceduralization to embed symbolic planning guidance directly into LM-based embodied agents, enabling efficient, adaptive reasoning in dynamic environments. By learning a vector-quantized procedural memory from symbolic plans and employing contrastive planning at test time, NeSyPr converts multi-step symbolic reasoning into single-step LM inference, eliminating the need for online symbolic tools. Empirical results across PDDLGym, VirtualHome, and ALFWorld show substantial gains in structured, adaptive, and timely reasoning while using compact LMs and achieving major latency reductions. The approach offers a scalable path to deploying embodied agents in latency-constrained settings, though its effectiveness hinges on the base LM and warrants future work on joint distillation and memory design improvements.

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.

Figures (11)

• Figure 1: The framework architecture of $\textsc{NeSyPr}$
• Figure 2: Vector-quantized procedural memory and contrastive planning
• Figure 3: Comparison with symbolic planner
• Figure 4: Analysis on procedural memory interpretability
• Figure 5: Application with different LMs