NeSyFS: A Neuro-symbolic Fast-Slow Thinking Framework for LLM Agent under Partial Observability

2026-07-30

AI Summary

This work addresses key challenges faced by large language model agents in partially observable environments—namely, difficulties in belief inference, goal misalignment, and planning under uncertainty—by proposing a neuro-symbolic fast-and-slow thinking framework inspired by human cognition. The fast system executes reactive actions, while the slow system maintains belief states through knowledge graph representations and performs uncertainty-aware planning via an improved twisted sequential Monte Carlo (TSMC) algorithm, complemented by a reflection mechanism to correct goal drift. This study is the first to integrate neuro-symbolic methods with dual-process reasoning, introducing knowledge graph–driven belief representation and a task-reflection–triggered strategy for switching between thinking modes. Evaluated on ALFWorld, WebShop, and ScienceWorld benchmarks, the approach significantly outperforms existing methods, demonstrating its effectiveness in complex partially observable tasks.

Abstract

Recently Large Language Models (LLMs) have been increasingly deployed as autonomous agents in applications such as self-reflection, retrieval-augmented generation, and scientific discovery. In these settings, agents must act based on limited observations rather than full environmental states, leading to partial observability. This introduces several key challenges: belief state inference, task objective misalignment, and planning under uncertainty. Prior approaches typically condition actions on full or summarized action-observation histories whose redundant and irrelevant information can mislead the decision making of LLM agent. Inspired by human cognition, we propose a novel neuro-symbolic fast-slow thinking (NeSyFS) framework for LLM agent, addressing the challenges introduced by partial observability in a unified approach. We use a knowledge graph (KG) to represent the belief state, providing triplets as context for every module of NeSyFS. The fast-thinking module performs reactive action, while slow-thinking conducts a new uncertainty-aware planning by following the high-level structure of twisted sequential Monte Carlo (TSMC) algorithm. To mitigate the misalignment of task objective, a reflection module is used to reflect fast-thinking actions, and also switches to the slow-thinking module whenever reactive actions repeatedly fail. Experiments on three representative benchmarks, i.e. ALFWorld, Webshop, and ScienceWorld, demonstrate significant advantages over previous methods.

Problem

• partial observability
• belief state inference
• task objective misalignment
• planning under uncertainty
• LLM agent

Innovation

• neuro-symbolic
• fast-slow thinking
• partial observability
• knowledge graph
• uncertainty-aware planning