Predicate Importance Estimation and Decoupled Rationale-Score Distillation for Entity Alignment

2026-06-22

AI Summary

This work addresses the challenge of semantic matching in knowledge graph entity alignment caused by predicate name discrepancies and incomplete local neighborhoods. To this end, the authors propose a Predicate Importance Estimation (PIE) module that generates more robust entity embeddings through predicate-aware weighted aggregation. Furthermore, they introduce a Decoupled Reasoning–Confidence Distillation (DRSD) framework, which leverages pseudo-labels generated by large language models to perform knowledge distillation in smaller models, explicitly separating reasoning rationale from confidence estimation. The proposed approach significantly improves alignment accuracy and enables the identification of high-risk samples through inconsistency between predicted labels and confidence scores, thereby facilitating a synergistic mechanism that combines automated processing with human verification.

Abstract

Knowledge graphs (KGs) are increasingly used as structured context for Large Language Models (LLMs), but industrial KG-RAG systems often need to integrate public and domain-specific KGs constructed from heterogeneous databases. This integration relies on Entity Alignment (EA), where lexical matching alone is insufficient under predicate-name variation and incomplete local neighborhoods. We address EA for KG integration by constructing a pairwise EA dataset and proposing two complementary modules: Predicate Importance Estimation (PIE) and Decoupled Rationale-Score Distillation (DRSD). PIE is a compact embedding-based approach that removes the subject information from each 1-hop triple, encodes the resulting subjectless triples, and aggregates them with learnable predicate-importance weights to build predicate-aware entity embeddings. DRSD trains a distilled small language model (SLM) with pseudo-answers produced by a teacher LLM through distinct prompts. By converting binary EA labels into text-based supervision and decoupling confidence-score estimation from label-consistent rationales, DRSD enables the SLM to learn task-specific reasoning while retaining a less label-biased confidence signal. Experiments show that PIE and DRSD improve EA classification. Moreover, because DRSD decouples confidence-score estimation from the decision, a discrepancy between the two flags an uncertain prediction for human review, thereby enabling a practical discrepancy between automatic acceptance and human-in-the-loop verification.

Problem

• Entity Alignment
• Knowledge Graph Integration
• Predicate Variation
• Incomplete Neighborhood
• KG-RAG

Innovation

• Predicate Importance Estimation
• Decoupled Rationale-Score Distillation
• Entity Alignment
• Knowledge Graph Integration
• Confidence-Rationale Decoupling