2026· Annual Meeting of the Association for Computational Linguistics· pp. 31190-31207· 0 citations· 42 references
Computer Science
TL;DR
An LLM reasoning framework with hierarchical relational retrieval for large-scale knowledge updating, named G-HiRel, which achieves superiority in terms of accuracy and interpretability and handles the knowledge inconsistency between the KG and LLM to obtain entity independence.
Abstract
Large language models (LLMs) have achieved good performance in multiple reasoning tasks. However, they are limited to adapt the rapid knowledge updates in the real-world scenario without retraining the entire LLM or modifying the model weights. Excluding these consuming methods, knowledge graphs (KGs) are used as external memory under knowledge updating because of their structural knowledge and efficient updating ability, which is yet limited by the gap between structural KG and LLM, and the deficient entity-independent semantics. To this end, we propose an LLM reasoning framework with hierarchical relational retrieval for large-scale knowledge updating, named G-HiRel. To integrate the structural edited KG into continuous LLMs, G-HiRel generates hierarchical instructions based on natural language questions. In order to handle the knowledge inconsistency between the KG and LLM and obtain the entity independence, G-HiRel utilizes a designed hierarchical relational retrieval for relational path candidates, which are selected by a designed semantics-based strategy. Finally, top entity-independent relational paths are in-stantiated and integrated into LLMs to generate the answer, in order to verify the reasoning performance under knowledge edits. Extensive experiments of G-HiRel on three benchmarks show that G-HiRel achieves superiority in terms of accuracy and interpretability. The code of G-HiRel is available at the link: https:
Experiments on three different domain tasks show that FKGLM can effectively integrate LLMs and large-scale knowledge graphs, leading to a significant enhancement in the reasoning capabilities of LLMs.
Yulin Zhou, Yongbin Qin, Chuan Lin· Journal of King Saud Univers...· 0 citations
This work proposes MARS, a scalable knowledge graph question answering (KGQA) approach that requires no model fine-tuning, and performs a structured retrieval procedure that links question entities to the KG and iteratively retrieves relevant next-hop information.
Nikit Srivastava, Daniel Vollmers, René Speck et al.· 0 citations
The externalisation of tacit knowledge, defined as practical, experiential knowledge that is difficult to formalise, represents one of the greatest challenges faced by institutions that rely on the expertise of their members. Due to its subjective and contextual nature, this type of knowledge resists capture through traditional methods such as interviews and observations, which are costly, difficult to standardise, and limited in reusability. This work proposes an architecture called K-GRASP (Knowledge Graph-based Retrieval-Augmented Structured Prompting), which combines the representational power of knowledge graphs (KG) with the probabilistic reasoning capabilities of Large Language Models (LLMs) to address this challenge. The proposed solution is divided into two phases: (i) a capture phase, in which an interviewer LLM conducts structured sessions with experts and converts their responses into RDF triples stored in a semantic graph, and (ii) a retrieval phase, in which a consultative LLM uses Retrieval-Augmented Generation (RAG) to translate natural language questions into SPARQL queries (SPARQL Protocol and RDF Query Language), retrieve relevant subgraphs, and generate contextualized responses. The architecture enables the systematic, reusable, and accessible codification of tacit knowledge, allowing for its large-scale preservation and dissemination. By integrating LLMs with formal representation structures, K-GRASP offers a robust, scalable, and interpretable solution to a historically complex problem in knowledge management. In the capture phase, prompts can be shaped to elicit concrete cases, boundary conditions, and decision rationales. At the same time, responses are mapped to an agreed-upon vocabulary to reduce drift across sessions. The graph can also retain provenance and scope cues (e.g., source expert, date, and stated assumptions), which may support later review and incremental refinement. In the retrieval phase, returning both the synthesised answer and the underlying triples can make the consultation more transparent and highlight gaps or ambiguities, as well as operational concerns such as access control, privacy, and versioning.
Rafael Luna, Gabriel S. Luna, C. E. Barbosa et al.· European Conference on Knowl...· 0 citations
The increasing deployment of large language models (LLMs) in enterprise environments creates reliability challenges related to hallucination, factual inconsistency, limited knowledge traceability, uncertainty, and operational efficiency. This study develops a literature-based architectural framework for reliable knowledge retrieval systems that separates external knowledge management from LLM-based reasoning and generation. The framework synthesizes architectural mechanisms for knowledge representation, hybrid retrieval, reranking, evidence selection, context construction, response verification, provenance tracking, uncertainty handling, guardrails, and computational efficiency. The resulting architecture organizes these mechanisms into coordinated layers that control the flow of external evidence from knowledge sources to generated responses while supporting traceability and controlled abstention when sufficient evidence is unavailable. The architectural synthesis further identifies complementary strategies for enterprise deployment, including semantic caching, model routing, and human oversight, to balance reliability, scalability, and operational cost. The analysis indicates that reliable LLM deployment should be treated as an end-to-end architectural problem rather than solely a model-performance problem, with knowledge access, evidence quality, verification, provenance, and governance functioning as integrated system components. The proposed framework provides a structured foundation for designing maintainable, auditable, and reliable knowledge retrieval systems for enterprise and other high-stakes applications
Bharat Kumar Reddy Karumuri· International Journal of Eng...· 0 citations
Large language models (LLMs) achieve remarkable performance across natural language tasks, yet they are trained on static corpora and their knowledge quickly becomes outdated in a fast-changing world. This motivates knowledge editing (KE), which updates specific knowledge in an LLM without changing unrelated others. Recent works move from structured knowledge triples toward unstructured KE (UKE), where the edit is a free-form passage that may state multiple facts at once. Nonetheless, existing editors inject such a passage yet fail to use it: the edited model can recall the passage, but can neither answer atomic questions about its facts nor compose them into multi-hop reasoning. We attribute this missing property, which we term composability, to editors'passive reliance on the fixed passage as the sole learning source. In response, we cast editing as a proactive self-distillation from a privileged in-context state of the same model, which requires no external supervision. We further reveal that due to the novelty of the injected knowledge, the pre-edited model's own rollouts rarely cover it, which limits the effectiveness of pure on-policy distillation. To close this gap, we propose HPSE, which builds a hybrid rollout that steps in to place missing facts onto the student's own trajectory precisely where its coverage fails, while staying on-policy elsewhere. We theoretically analyze HPSE's advantage over pure on-policy distillation, and empirically establish its plug-and-play improvements across four LLM backbones and two KE editors under various scenarios.
Tianci Liu, Zihan Dong, Tianchun Li et al.· 0 citations
Retrieval-augmented methods improve factual accuracy by grounding language models in external knowledge, but retrieving over unstructured text often introduces irrelevant context and offers limited control over the retrieved information. Structured knowledge bases offer a more controllable alternative, yet they are expensive to construct and often brittle to reason over. To address these limitations, we propose KBevo: a co-evolving framework that jointly learns to construct a structured knowledge base and reason over it for knowledge-intensive question answering. By optimizing both components end-to-end with QA outcome rewards, our method enables reasoning success to directly improve the quality of the constructed knowledge base. This leads to larger, better-connected knowledge structures with higher answer reachability, while also improving compositional factual reasoning and controllability compared to standard retrieval baselines.
Ryan Thomas Noonan, Lin-Xi Zhao, Meng-Han Xu et al.· 0 citations