Tokenization, Fusion and Decoupling: Bridging the Granularity Mismatch Between Large Language Models and Knowledge Graphs

Siyue Su; Jian Yang; Bo Li; Guanglin Niu

cs.CLFeb 26, 2026

Tokenization, Fusion and Decoupling: Bridging the Granularity Mismatch Between Large Language Models and Knowledge Graphs

Authors: Siyue Su, Jian Yang, Bo Li, Guanglin Niu

Organizations: Beihang University

Abstract

Leveraging Large Language Models (LLMs) for Knowledge Graph Completion (KGC) is promising but hindered by a fundamental granularity mismatch. LLMs operate on fragmented token sequences, whereas entities are the fundamental units in knowledge graphs (KGs) scenarios. Existing approaches typically constrain predictions to limited candidate sets or align entities with the LLM's vocabulary by pooling multiple tokens or decomposing entities into fixed-length token sequences, which fail to capture both the semantic meaning of the text and the structural integrity of the graph. To address this, we propose KGT, a novel framework that uses dedicated entity tokens to enable efficient, full-space prediction. Specifically, we first introduce specialized tokenization to construct feature representations at the level of dedicated entity tokens. We then fuse pre-trained structural and textual features into these unified embeddings via a relation-guided gating mechanism, avoiding training from scratch. Finally, we implement decoupled prediction by leveraging independent heads to separate and combine semantic and structural reasoning. Experimental results show that KGT consistently outperforms state-of-the-art methods across multiple benchmarks.

Figures & tables

Figure 1: An illustration of existing two strategies of full-space LLM-based methods and KGT. (a) Pooling multiple tokens to unified representations for entities; (b) Decomposing entities into fixed-length sub-word sequences; (c) Constructing feature representations directly at the indivisible entity level.

Figure 2: Overview of the KGT framework. Part 1 illustrates the overall pipeline of KGT. The tokenizer first processes the input text containing the incomplete triple query, where entities and relations are represented as special tokens added to the original vocabulary. These special tokens obtain their embeddings via the Dual-Stream Specialized Token Embedding module. Subsequently, the LLM Backbone encodes the sequence, extracting the feature of the last token, which is then fed into the Dual-View Decoupled Predictor to generate the probability distribution over the entire entity vocabulary. Part 2 details the implementation of the Dual-Stream Specialized Token Embedding, where the dashed line indicates the assignment of the fused specialized feature to the special token representing the head entity h. Part 3 depicts the detailed architecture of the Dual-View Decoupled Predictor.

Methods	MKG-W		MKG-Y		DB15K
Methods	MRR $\uparrow$	H@1 $\uparrow$	MRR $\uparrow$	H@1 $\uparrow$	MRR $\uparrow$	H@1 $\uparrow$	H@3 $\uparrow$	H@10 $\uparrow$
TransE	29.19	21.06	30.73	23.45	24.86	12.78	31.48	47.07
DistMult	20.99	15.93	25.04	19.33	23.03	14.78	26.28	39.59
RotatE	33.67	26.80	34.95	29.10	29.28	17.87	36.12	49.66
TuckER	30.39	24.44	37.05	34.59	33.86	25.33	37.91	50.38
IKRL	32.36	26.11	33.22	30.37	26.82	14.09	34.93	49.09

Table 1: The main KGC results.

\clubsuit

represents the experimental results that we reproduced through source code. The best and second-best results are boldfaced and underlined, respectively. -: unavailable entry.

Setting		MKG-W		MKG-Y		DB15K
Setting		MRR	H@1	MRR	H@1	MRR	H@1
Full Model		43.27	36.02	43.61	37.68	42.16	34.06
Modality Contribution	(1.1). Structure Modality	32.73	27.16	36.12	33.39	37.62	28.85
Modality Contribution	(1.2). Text Modality	40.69	32.44	40.86	34.36	38.08	30.14
Model Design	(2.1). w/o structural input	41.43	33.31	42.15	35.91	41.57	33.47
	(2.2). w/o textual input	39.90	32.78	39.10	33.78	41.74	33.53

Table 2: Ablation study of different settings on MKG-W, MKG-Y, and DB15K datasets.

Figure 3: A comprhensive comparison between several viriants of KGT on DB15K.

Figure 4: Trainable parameters of some LLM-based KGC methods based on MKG-W.

Figure 5: Results of different logits scaling.

Figure 6: KGC results using different feature extractors. We evaluate diverse textual feature extractors ( text-embedding-3-large , bge-large-en-v1.5 ) on the MKG-W dataset, and structural feature extractors (TransE, RotatE) on the DB15K dataset.

Figure 7: The comparison of model weights among various cases

Case 1: ( ? , composer , Adaptation_(film) )
Textual	Rank of the head entity “ Carter Burwell ” : 179
Structural	Rank of the head entity “ Carter Burwell ” : 1
KGT	Rank of the head entity “ Carter Burwell ” : 1
Case 2: ( ? , distributingCompany , Warner_Music_Group )
Textual	Rank of the head entity “ Atlantic_Records ” : 1
Structural	Rank of the head entity “ Atlantic_Records ” : 29

Table 3: Case study experiments on the DB15K dataset

Appendix figures & tables5 assets

Supplementary material from the paper’s appendix.

Appendix

Dataset	LLM	text dim	struct dim	LoRA $r$	LoRA $\alpha$	LORA target modules	logits scaling ratio $\gamma$
MKG-W	Llama-2-7b	1536	256	8	16	query, value	1.4
MKG-Y	Llama-2-7b	1536	256	8	16	query, value	1.6
DB15K	Llama-2-7b	1536	256	8	16	query, value	1.5
Dataset	epoch	batch size per device	gradient accumulation steps	learning rate	optimizer	text dropout	struct dropout
MKG-W	8	32	1	2e-4	Adamw_8bit	0.2	0.4
MKG-Y	8	32	1	2e-4	Adamw_8bit	0.1	0.3

Appendix

Table 4: Hyper-parameter settings in the main experients.

Suppose that you are an excellent linguist studying a new three-word language for knowledge graph. Given the following dictionary:

Input Type Description

<

kgl: Mainz

>

Head entity Mainz

<

kgl: capital of

>

Relation capital of

Please complete the last word (?) of the sentence:

<

kgl: Mainz

>

<

kgl: capital of

>

Table 5: The prompt of KGT. Note that we take entity "Mainz" and relation "capital of" as a example.

Appendix

Table 5: The prompt of KGT. Note that we take entity "Mainz" and relation "capital of" as a example.

Dataset	#Entity	#Relation	#Train	#Valid	#Test	Text
MKG-W	15000	169	34196	4276	4274	14123
MKG-Y	15000	28	21310	2665	2663	12305
DB15K	12842	279	79222	9902	9904	9078

Appendix

Table 6: Statistical information of the three datasets in our experiments. The entity descriptions are provided by the original datasets.

Methods	MKG-W				MKG-Y
Methods	MRR	H@1	H@3	H@10	MRR	H@1	H@3	H@10
TransE	29.19	21.06	33.20	44.23	30.73	23.45	35.18	43.37
RotatE	33.67	26.80	36.68	46.76	34.95	29.10	38.35	45.30
TuckER	30.39	24.44	32.91	41.25	37.05	34.59	38.43	41.45
IKRL	32.36	26.11	34.75	44.07	33.22	30.37	34.28	38.60
TransAE	30.00	21.23	34.91	44.72	28.10	25.31	29.10	33.03

Appendix

Table 7: Full results on the MKG-W and MKG-Y datasets.

Methods	WN18RR
Methods	MRR $\uparrow$	Hits@1 $\uparrow$	Hits@3 $\uparrow$	Hits@10 $\uparrow$
TransE	.232	.061	.366	.522
RotatE	.476	.428	.492	.571
TuckER	.470	.443	.526	.526
KG-BERT	.216	.041	.302	.524
StAR	.401	.243	.491	.709

Appendix

Table 8: Additional results on WN18RR.

KG Completion

Knowledge graph completion (KGC) predicts missing entities, relations, or facts from structured graphs and supporting evidence.

See all 76 papers

GS-Quant: Granular Semantic and Generative Structural Quantization for Knowledge Graph Completion
Apr 23, 2026Qizhuo Xie, Yunhui Liu, Yu Xing +4Graph Representation Learning KG Completion
State Key Laboratory for Novel Software Technology, Nanjing University, China
OMNIA: Closing the Loop by Leveraging LLMs for Knowledge Graph Completion
Mar 12, 2026Frédéric Ieng, Soror Sahri, Mourad Ouzzani +5KG Completion LLMs for Graph Learning
LIPADE, Université Paris Cité, Paris, France · Qatar Computing Research Institute, HBKU, Doha, Qatar · Universität zu Lübeck, Germany +3
The Graph Language: How Knowledge Graphs Speak to Large Language Models
Aug 2, 2026Giuseppe PirròLLM Grounding Knowledge Augmentation for Language Models
DeMaCS, University of Calabria

Apr 23, 2026J/K move · Enter open · S save

GS-Quant: Granular Semantic and Generative Structural Quantization for Knowledge Graph Completion

Qizhuo Xie, Yunhui Liu, Yu Xing, Qianzi Hou+3

State Key Laboratory for Novel Software Technology, Nanjing University, China

Graph Representation Learning KG Completion

Open paper AskPDF arXiv

Large Language Models (LLMs) have shown immense potential in Knowledge Graph Completion (KGC), yet bridging the modality gap between continuous graph embeddings and discrete LLM tokens remains a critical challenge. While recent quantization-based approaches attempt to align these modalities, they typically treat quantization as flat numerical compression, resulting in semantically entangled codes that fail to mirror the hierarchical nature of human reasoning. In this paper, we propose GS-Quant, a novel framework that generates semantically coherent and structurally stratified discrete codes for KG entities. Unlike prior methods, GS-Quant is grounded in the insight that entity representations should follow a linguistic coarse-to-fine logic. We introduce a Granular Semantic Enhancement module that injects hierarchical knowledge into the codebook, ensuring that earlier codes capture global semantic categories while later codes refine specific attributes. Furthermore, a Generative Structural Reconstruction module imposes causal dependencies on the code sequence, transforming independent discrete units into structured semantic descriptors. By expanding the LLM vocabulary with these learned codes, we enable the model to reason over graph structures isomorphically to natural language generation. Experimental results demonstrate that GS-Quant significantly outperforms existing text-based and embedding-based baselines. Our code is publicly available at https://github.com/mikumifa/GS-Quant.

Abstract

KG Completion

Explore similar work

GS-Quant: Granular Semantic and Generative Structural Quantization for Knowledge Graph Completion

OMNIA: Closing the Loop by Leveraging LLMs for Knowledge Graph Completion

The Graph Language: How Knowledge Graphs Speak to Large Language Models