RADAR: Reasoning as Discrimination with Aligned Representations for LLM-based Knowledge Graph Reasoning

Bo Xue; Yuan Jin; Luoyi Fu; Jiaxin Ding; Xinbing Wang

cs.CLFeb 25, 2026

RADAR: Reasoning as Discrimination with Aligned Representations for LLM-based Knowledge Graph Reasoning

Authors: Bo Xue, Yuan Jin, Luoyi Fu, Jiaxin Ding, Xinbing Wang

Organizations: Shanghai Jiao Tong University, Shanghai, China

Abstract

Knowledge graph reasoning (KGR) infers missing facts, with recent advances increasingly harnessing the semantic priors and reasoning abilities of Large Language Models (LLMs). However, prevailing generative paradigms are prone to memorizing surface-level co-occurrences rather than learning genuine relational semantics, limiting out-of-distribution generalization. To address this, we propose RADAR, which reformulates KGR from generative pattern matching to discriminative relational reasoning. We recast KGR as discriminative entity selection, where reinforcement learning enforces relative entity separability beyond token-likelihood imitation. Leveraging this separability, inference operates directly in representation space, ensuring consistency with the discriminative optimization and bypassing generation-induced hallucinations. Across four benchmarks, RADAR achieves 5-6% relative gains on link prediction and triple classification over strong LLM baselines, while increasing task-relevant mutual information in intermediate representations by 62.9%, indicating more robust and transferable relational reasoning.

Figures & tables

Figure 1: FB15K-237N triple classification with a shared LLaMA backbone: co-occurrence-based SFT yields only marginal gains over the frozen baseline, whereas RADAR substantially improves accuracy by optimizing discriminative relational reasoning.

Figure 2: RADAR, a data-to-inference alignment framework for discriminative KGR.

Figure 3: Training prompt.

Figure 4: Inference prompt.

Paradigm	Model	FB15K-237				FB15K-237N
Paradigm	Model	MRR	H@1	H@3	H@10	MRR	H@1	H@3	H@10
Embedding -based	TransE	0.279	0.198	0.376	0.441	0.255	0.152	0.301	0.459
	DistMult	0.281	0.199	0.301	0.446	0.209	0.143	0.234	0.330
	ComplEx	0.278	0.194	0.297	0.450	0.249	0.180	0.276	0.380
	RotatE	0.338	0.241	0.375	0.533	0.279	0.177	0.320	0.481
	ConvE	0.312	0.225	0.341	0.497	0.273	0.192	0.305	0.429

Table 1: Link prediction performance (MRR and Hits@k) on FB15K-237 and FB15K-237N. Best scores are in bold and second-best are underlined .

Paradigm	Model	WN18RR	FB15K-237N	UMLS
Embedding -based	TransE	88.4	69.7	84.5
	DistMult	85.1	58.7	86.4
	ComplEx	84.1	65.7	87.1
	RotatE	88.2	68.5	87.7
Language Model -based	KG-BERT	91.6	56.0	77.3
	LLAMA-ICL	50.2	59.2	55.5

Table 2: Triple classification accuracy on WN18RR, FB15K-237N, and UMLS. Best scores are in bold and second-best are underlined .

Answer Cardinality		Negative Hardness
(Fixing Negatives: Tier 2)		(Fixing Cardinality: Variable)
Setting	Accuracy	Setting	Accuracy
Single-answer	79.9	Tier 1 (Easy)	80.2
Variable-answer	81.6	Tier 2 (Medium)	81.6
		Tier 3 (Hard)	81.0

Table 3: Triple classification accuracy on FB15K-237N under controlled ablations of task difficulty: answer cardinality and negative hardness.

Method	WN18RR	15K-237N	UMLS
Serial-SFT-Generation	0.921	0.748	0.858
Serial-SFT-Extraction	0.925	0.753	0.861
Discriminative-SFT-Extraction	0.945	0.793	0.898
Discriminative-Full-Extraction	0.953	0.816	0.917

Table 4: Ablation across task formulation, training objective, and inference strategy on triple classification accuracy (WN18RR, FB15k-237N, UMLS).

Figure 5: Task-Adaptive SMI of intermediate representations across different methods on WN18RR.

LLaMA-7B Touvron et al. (2023)
Method	MRR	H@1	H@3	H@10
Serial-SFT-Extraction	0.365	0.229	0.438	0.584
Discriminative-Full-Extraction	0.415	0.301	0.476	0.633
Pythia-6.9B Biderman et al. (2023)
Serial-SFT-Extraction	0.353	0.259	0.401	0.616
Discriminative-Full-Extraction	0.384	0.267	0.437	0.635

Table 5: Link prediction performance on FB15K-237N across different LLM backbones.

Figure 6: Inductive generalization performance for triple classification on WN18RR under varying inductive rates (IR). We report accuracy on seen (S), unseen (U), and overall (A) test triples.

Subjects	w/o Training	w/ Training
Clinical	0.340	0.455
College Medicine	0.341	0.397
High School Biology	0.313	0.439
High School Chemistry	0.271	0.355
Medical Genetics	0.290	0.490

Table 6: Zero-shot performance on domain-aligned MMLU subjects after UMLS-based KGR training.

Appendix figures & tables6 assets

Supplementary material from the paper’s appendix.

Appendix

Dataset	$\|\mathbf{\mathcal{E}}\|$	$\|\mathbf{\mathcal{R}}\|$	# Train	# Valid	# Test
FB15K-237	14,541	237	272,115	17,535	20,466
FB15K-237N	13,104	93	87,282	7,041	8,226
WN18RR	40,943	11	86,835	6,068	6,268
UMLS	135	46	5,216	1,304	1,322

Appendix

Table 7: Dataset statistics.

Parameter $n$	MRR
15	0.415
20	0.402
25	0.393

Appendix

Table 8: Link classification performance on FB15K-237N under varying parameter

n

Figure 7: Task-Adaptive SMI of intermediate representations across different methods on FB15K-237N.

Figure 8: COT prompt.

Figure 9: Layer-wise triple classification accuracy of RADAR across datasets. The horizontal axis denotes the layer depth in LLaMA, and the vertical axis corresponds to different datasets.

Parameter $K$	Accuracy
3	0.892
4	0.898
5	0.887
6	0.893

Appendix

Table 9: Triple classification accuracy on UMLS under varying parameter

K

of Discriminative-SFT-Extraction. The best result is highlighted in bold.

Graph Reasoning

Graph reasoning uses relationships among entities to solve multi-step inference, prediction, planning, and question-answering tasks.

See all 62 papers

Structure-Internalized Rule Language Model for Faithful Knowledge Graph Reasoning
Aug 18, 2026Xingrui Zhuo, Jiapu Wang, Manzong Huang +2LLM Reasoning with Graphs Logical Reasoning
Key Laboratory of Knowledge Engineering with Big Data (Hefei University of Technology), Ministry of Education, China · School of Computer Science and Information Engineering, Hefei University of Technology, China · Nanjing University of Science and Technology, China
KGFR: A Foundation Retriever for Generalized Knowledge Graph Question Answering
Nov 6, 2025Yuanning Cui, Zequn Sun, Wei Hu +1Multi-Hop QA LLM Grounding
School of Computer Science, Nanjing University of Information Science and Technology, Nanjing 210044, China · State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing 210023, China · National Institute of Healthcare Data Science, Nanjing University, Nanjing 210093, China +2
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

Aug 18, 2026J/K move · Enter open · S save

Structure-Internalized Rule Language Model for Faithful Knowledge Graph Reasoning

Xingrui Zhuo, Jiapu Wang, Manzong Huang, Gongqing Wu+1

Key Laboratory of Knowledge Engineering with Big Data (Hefei University of Technology), Ministry of Education, China · School of Computer Science and Information Engineering, Hefei University of Technology, China · Nanjing University of Science and Technology, China

LLM Reasoning with Graphs Logical Reasoning Knowledge Graph Question Answering KG Completion Neuro-Symbolic Reasoning

Open paper AskPDF arXiv

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.

Abstract

Graph Reasoning

Explore similar work

Structure-Internalized Rule Language Model for Faithful Knowledge Graph Reasoning

KGFR: A Foundation Retriever for Generalized Knowledge Graph Question Answering

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