Advancing Symbolic Integration in Large Language Models: Beyond Conventional Neurosymbolic AI

TL;DR

The paper surveys symbolic integration with Large Language Models (LLMs) to address transparency and reasoning in high-risk domains. It outlines a four-dimensional roadmap—integration stages, coupling mechanisms, architectural paradigms, and application/algorithm-level perspectives—grounded in a structured literature review of 2018–Feb 2025 that identifies benchmarks and gaps. It distinguishes NeSy AI as a broader precursor to Symbolic-integrated LLMs, arguing for specialized adaptation to LLMs, and then details architectural paradigms (LLM-to-Symbolic, Symbolic-to-LLM, and Hybrid) with concrete mechanisms (KGs, logic, prompts, adapters, and RAG) and extensive benchmark discussions. The work highlights state-of-the-art achievements such as ground-grounding via GraphRAG, LLM-ARC, and logic-enhanced inference, while candidly outlining persistent challenges—design patterns, evaluation standards, knowledge updates, and efficiency—providing a practical roadmap for future research and development in transparent, trustworthy AI systems.

Abstract

LLMs have demonstrated highly effective learning, human-like response generation,and decision-making capabilities in high-risk sectors. However, these models remain black boxes because they struggle to ensure transparency in responses. The literature has explored numerous approaches to address transparency challenges in LLMs, including Neurosymbolic AI (NeSy AI). NeSy AI approaches were primarily developed for conventional neural networks and are not well-suited to the unique features of LLMs. Consequently, there is a limited systematic understanding of how symbolic AI can be effectively integrated into LLMs. This paper aims to address this gap by first reviewing established NeSy AI methods and then proposing a novel taxonomy of symbolic integration in LLMs, along with a roadmap to merge symbolic techniques with LLMs. The roadmap introduces a new categorisation framework across four dimensions by organising existing literature within these categories. These include symbolic integration across various stages of LLM, coupling mechanisms, architectural paradigms, as well as algorithmic and application-level perspectives. The paper thoroughly identifies current benchmarks, cutting-edge advancements, and critical gaps within the field to propose a roadmap for future research. By highlighting the latest developments and notable gaps in the literature, it offers practical insights for implementing frameworks for symbolic integration into LLMs to enhance transparency.

Figures (10)

• Figure 1: Neurosymbolic AI architectures with different symbolic reasoning paradigms: (a) Knowledge Graph-Neural Network Integration; (b) Rule-Neural Network Integration and (c) Logic-Neural Network Integration. Each subfigure shows the information flow between symbolic and neural components with inputs and outputs at each processing stage.
• Figure 2: PRISMA flow diagram showing the systematic literature review process
• Figure 3: Distribution analysis of included studies in the systematic literature review: (a) database coverage, (b) temporal spread by year, (c) journal quality distribution across quartiles, and (d) conference quality distribution across tier groups.
• Figure 4: Roadmap for Symbolic Integration with Large Language Models (LLMs)
• Figure 5: Neurosymbolic Approaches depending on the interaction level