Aug 2026· Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2· 0 citations· 18 references
TL;DR
RGRS is proposed, a unified neuro-symbolic framework that integrates rule-level reasoning with graph representation learning for scalable Horn rule instantiation and achieves consistently high accuracy with significant efficiency gains over symbolic and neuro-symbolic baselines.
Abstract
Knowledge graph reasoning is a fundamental problem in artificial intelligence, where rule-based methods provide strong interpretability but suffer from exponential symbolic search complexity, while existing neuro-symbolic approaches struggle to model complex Horn rules with rich topologies. We propose Rule–Graph Joint Representation Learning (RGRS), a unified neuro-symbolic framework that integrates rule-level reasoning with graph representation learning for scalable Horn rule instantiation. RGRS introduces logically constrained rule embeddings and dominance-aware subgraph representations, enabling embedding-guided structural pruning while preserving correctness through exact symbolic verification. By coupling continuous structural abstraction with deterministic matching in a filter–verify pipeline, RGRS effectively reduces the combinatorial search space without compromising logical semantics. Extensive experiments on large-scale knowledge graphs show that RGRS achieves consistently high accuracy with significant efficiency gains over symbolic and neuro-symbolic baselines, providing a principled bridge between symbolic reasoning and embedding-based efficiency.
Large Language Models (LLMs) have demonstrated remarkable proficiency in natural language understanding, yet they struggle with strict multi-step reasoning, frequently suffering from hallucinations and inconsistency. Existing solutions like Chain-of-Thought (CoT) lack rigorous verification mechanisms, while standard Retrieval-Augmented Generation (RAG) often misses the complex, structural dependencies inherent in logical tasks. To bridge this gap, we propose a Neuro-Symbolic architecture that integrates a Logical Knowledge Graph (LKG) with dynamic solver routing. Specifically, we introduce an ontology-based LKG that treats logical rules and constraints as first-class topological nodes, enabling explicit modeling of dependencies extracted from text. We further design a Logic Router to dynamically dispatch tasks to the optimal symbolic engine, which is supported by a topology-aware hybrid retrieval mechanism. Experimental results on logical reasoning benchmarks demonstrate that our framework significantly outperforms state-of-the-art prompting and RAG baselines, delivering higher accuracy and verifiable reasoning paths.
Hai-Zhao Fan, Yu-Chi Xiong, Jize Wang et al.· 0 citations
Evaluated across nine SMT logics, SMT-Select consistently outperforms existing selectors and SMT-COMP winning solvers and closes at least 30% of the performance gap between the competition winner and the virtual best solver (VBS).
Zhengyang Lu, Paul Sarnighausen-Cahn, Jiahao Chen et al.· International Conference on...· 0 citations
Large language models (LLMs) often produce reasoning steps that are superficially coherent yet internally inconsistent, leading to unreliable outputs. Since such failures typically arise from implicit or poorly-grounded knowledge, we introduce Grounded Reasoning in Dependency (GRiD) , a novel dependency-aware reasoning framework that explicitly grounds reasoning steps in structured knowledge. GRiD represents reasoning as a graph consisting of interconnected knowledge extraction nodes and reasoning nodes, enforcing logical consistency through explicit dependencies. Each reasoning step is validated via a lightweight, step-wise verifier that ensures logical correctness relative to its premises. Extensive experiments across diverse reasoning benchmarks—including StrategyQA, CommonsenseQA, GPQA, and TruthfulQA—demonstrate that GRiD substantially improves reasoning accuracy, consistency, and faithfulness compared to recent state-of-the-art structured reasoning methods. Notably, GRiD enhances performance even when applied purely as a lightweight verification module at inference time, underscoring its generalizability and practical utility † .
Xiangyu Wen, Min Li, Junhua Huang et al.· Neural Information Processin...· 2 citations
A unified framework that embeds unstructured text into structured knowledge graphs, creating a heterogeneous network for flexible evidence retrieval, outperforming SOTA baselines in answer accuracy and reasoning fidelity while maintaining extremely low token costs and near real-time inference is proposed.
In machine reading comprehension (MRC) tasks that involve complex logical reasoning, graph structures are often regarded as a more suitable representation paradigm than plain text sequences, owing to their explicit structured semantic environment and strong relational inference capabilities. Nevertheless, most existing graph-based MRC models commonly suffer from incomplete node representations and sparse connectivity during graph construction. Moreover, these models typically devote most of their capacity to refining node features while comparatively overlooking the rich relational information encoded in edges. Such limitations significantly curb the potential of graph structures to support deep relational understanding and complex reasoning tasks. In light of this, this study proposes a Graph-Structured Collaborative Inference (GSCI) framework for the task of logical reading comprehension, which consists of three key modules: LLM-empowered logic-oriented graph construction, dual-stream collaborative graph modeling, and structure-aware answer prediction. Firstly, GSCI leverages an LLM to perform fine-grained semantic parsing, extracting abundant structured knowledge that constructs logic-oriented graph foundation for downstream inference. Secondly, a node–edge cooperative modeling mechanism is designed to jointly capture semantic information from nodes and relational signals from edges. Thirdly, the learned structured representations are integrated with textual representations to support candidate answer prediction. Finally, comprehensive experiments on two challenging logical reasoning benchmarks (ReClor and LogiQA) demonstrate that GSCI consistently outperforms existing baselines, providing strong empirical evidence of its effectiveness in logical reading comprehension tasks.
Zhuo Zhao, Zhiwen Xie, Guangyou Zhou et al.· IEEE Transactions on Audio,...· 0 citations
Knowledge Graph Reasoning (KGR) aims to discover latent facts by leveraging the structural evidence available in KGs, posing a challenge to the structural semantic understanding capability of KGR models. Recent studies have demonstrated that Large Language Models (LLMs) can achieve remarkable progress on KGR tasks via flexible in-context learning. However, the inherent representation inconsistency between KG structural context and LLM parametric knowledge remains inadequately addressed. This limitation prevents LLMs from effectively perceiving reasoning evidence that aligns with KG constraints, which undermines both the effectiveness and faithfulness of reasoning. We refer to this problem as reasoning evidence perception drift of LLMs over KGs. To address this problem, we propose a Structure-Internalized Rule Language Model (SIRLM), which centers on structural rule generation to couple the parametric learning of structural knowledge with the faithfulness evaluation of reasoning logic, enabling LLMs to anchor tightly to KG-grounded evidence. Specifically, we first design a Structure-Internalized Rule Generator (SIRG), which incorporates an in-context learning block augmented with a structural relation memory to coordinate structural and parametric knowledge. Furthermore, we equip SIRG with a KG tokenizer based on structural invariance learning and a neuro-symbolic reasoner based on rule-constrained message propagation. These components provide SIRG with learnable structural representations and faithful rule-execution feedback, respectively. Our SIRLM can be seamlessly integrated into standard LLM training paradigms, such as SFT and GRPO. Extensive experiments against 17 state-of-the-art KGR methods on 36 datasets demonstrate the significant superiority of SIRLM.
Xingrui Zhuo, Jiapu Wang, Manzong Huang et al.· 0 citations