arXiv:2510.09711v2 Announce Type: replace 
Abstract: Large Language Models (LLMs) have recently emerged as a powerful paradigm for Knowledge Graph Completion (KGC), offering strong reasoning and generalization capabilities beyond traditional embedding-based approaches. However, existing LLM-based methods often struggle to fully exploit structured semantic representations, as the continuous embedding space of pretrained KG models is fundamentally misaligned with the discrete token space of LLMs. This discrepancy hinders effective semantic transfer and limits their performance. To address this challenge, we propose ReaLM, a novel and effective framework that bridges the gap between KG embeddings and LLM tokenization through the mechanism of residual vector quantization. ReaLM discretizes pretrained KG embeddings into compact code sequences and integrates them as learnable tokens within the LLM vocabulary, enabling seamless fusion of symbolic and contextual knowledge. Furthermore, we incorporate ontology-guided class constraints to enforce semantic consistency, refining entity predictions based on class-level compatibility. Extensive experiments on two widely used benchmark datasets demonstrate that ReaLM achieves state-of-the-art performance, confirming its effectiveness in aligning structured knowledge with large-scale language models.

يقدم إطار ReaLM الجديد حلاً للتحديات التي تواجه النماذج اللغوية الكبيرة (LLMs) في الاستفادة الفعالة من التمثيلات الدلالية الهيكلية من الرسوم البيانية المعرفية (KGs). من خلال استخدام الكوانتيزات المتبقية، يدمج ReaLM تمثيلات KG في توكين النماذج اللغوية الكبيرة، مما يسهل دمج المعرفة الرمزية والسياقية بشكل أكثر سلاسة.

La introducción de ReaLM, un nuevo marco, aborda los desafíos que enfrentan los grandes modelos de lenguaje (LLMs) para utilizar de manera efectiva las representaciones semánticas estructuradas de los grafos de conocimiento (KGs). Al emplear la cuantización vectorial residual, ReaLM integra los embeddings de KG en la tokenización de LLM, facilitando una fusión más fluida del conocimiento simbólico y contextual.

L'introduction de ReaLM, un nouveau cadre, répond aux défis rencontrés par les grands modèles de langage (LLMs) pour utiliser efficacement les représentations sémantiques structurées des graphes de connaissances (KGs). En utilisant la quantification vectorielle résiduelle, ReaLM intègre les embeddings de KG dans la tokenisation des LLM, facilitant une fusion plus fluide des connaissances symboliques et contextuelles.

The introduction of ReaLM, a novel framework, addresses the challenges faced by large language models (LLMs) in effectively utilizing structured semantic representations from knowledge graphs (KGs). By employing residual vector quantization, ReaLM integrates KG embeddings into LLM tokenization, facilitating a more seamless fusion of symbolic and contextual knowledge.