ReaORE: Reasoning-Guided Progressive Open Relation Extraction Empowered by Large Reasoning Models

2026-06-25

AI Summary

Open relation extraction (OpenRE) faces significant challenges in generalizing to unseen relations and distinguishing between easily confusable ones. To address these issues, this work proposes ReaORE, a framework that employs a two-stage progressive reasoning paradigm: it first generates a candidate relation set by integrating multi-dimensional semantic understanding with embedding similarity filtering, and then performs fine-grained comparative reasoning to accurately predict the target relation. ReaORE innovatively combines the reasoning capabilities of large language models, multi-dimensional relation modeling, and a contrastive mechanism, substantially enhancing generalization and discrimination for unseen relations. Experimental results demonstrate that ReaORE significantly outperforms existing baselines on two mainstream OpenRE benchmarks, confirming its effectiveness and robustness.

Abstract

Open Relation Extraction (OpenRE) requires a model to extract unseen relations between head and tail entities from unstructured text for real-world applications. The core challenge of OpenRE lies in achieving reliable generalization to unseen relation types. Current OpenRE approaches either employ clustering techniques, which cannot generate relation labels and suffer from poor generalization, or rely on direct relation label generation via Large Language Models (LLMs), which lack sufficient discriminative capacity to distinguish easily confused relations. To address these limitations, we propose Reasoning-guided progressive OpenRE (ReaORE), a framework for performing relation extraction through coarse-to-fine relation reasoning. Specifically, ReaORE consists of two key stages: (i) relation filtering, which reasons over multiple aspects to understand relations and instances, yielding an initial relation set, and further supplements and filters relations via embedding-based similarity to ensure the target relation is included; (ii) relation prediction, which aims to predict the target relations from the above set via fine-grained comparative reasoning to better distinguish easily confused relations. Extensive experiments on two widely used OpenRE datasets demonstrate that ReaORE outperforms existing baselines.

Problem

• Open Relation Extraction
• unseen relations
• relation generalization
• discriminative capacity
• relation confusion

Innovation

• Open Relation Extraction
• Reasoning-Guided Framework
• Coarse-to-Fine Reasoning
• Large Reasoning Models
• Relation Discrimination