HELEA: Hard-Negative Benchmark and LLM-based Reranking for Robust Entity Alignment

2026-05-27

AI Summary

This study addresses the vulnerability of existing knowledge graph entity alignment benchmarks to name overlap, which hinders robust evaluation of models in scenarios involving distinct entities with identical names. To this end, the authors introduce DW-HN29K, the first systematically constructed benchmark featuring hard negative examples with shared names. They propose a two-stage alignment framework: first, a one-hop context encoder trained on hard negatives enables efficient candidate retrieval; second, a large language model—used without fine-tuning—performs reranking. This approach substantially reduces reliance on entity names, achieving an F1 score of 0.967 on DW-HN29K while maintaining a Hit@1 of 0.993 on the standard DW-15K benchmark, significantly outperforming existing name-dependent methods.

Abstract

Entity Alignment (EA) is essential for knowledge graph (KG) fusion, but existing benchmarks often allow models to exploit name overlap rather than relational structure. This makes it difficult to evaluate whether models can reject same-name entities that refer to different real-world objects. Our primary contribution is a same-name hard-negative augmentation strategy that simultaneously yields quality-controlled evaluation benchmarks (DW-HN29K, DY-HN27K) and augmented training corpora (DW-Train, DY-Train), by mining same-name but distinct entity pairs from KG name-collision groups. We further introduce HELEA, a two-stage framework integrating (i) entity encoder retrieval trained on hard-negative-augmented training corpora with 1-hop KG context, and (ii) LLM-based reranking without additional training. Experiments show that name-dependent baselines collapse to near-random performance on our hard-negative benchmarks, while HELEA achieves F1 0.967 on DW-HN29K while maintaining Hit@1 0.993 on standard DW-15K.

Problem

• Entity Alignment
• Hard-Negative Benchmark
• Knowledge Graph Fusion
• Name Collision
• Robustness

Innovation

• Hard-Negative Mining
• Entity Alignment
• LLM-based Reranking
• Knowledge Graph Fusion
• Name Collision