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

TL;DR

RADAR is proposed, which reformulates KGR from generative pattern matching to discriminative relational reasoning, and recast KGR as discriminative entity selection, where reinforcement learning enforces relative entity separability beyond token-likelihood imitation.

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 (9)

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