Communication Dans Un Congrès Année : 2026

Link Prediction or Perdition: the Seeds of Instability in Knowledge Graph Embeddings

Résumé

Embedding models (KGEMs) constitute the main link prediction approach to complete knowledge graphs. Standard evaluation protocols emphasize rank-based metrics such as MRR or Hits@K, but usually overlook the influence of random seeds on result stability. Moreover, these metrics conceal potential instabilities in individual predictions and in the organization of embedding spaces. In this work, we conduct a systematic stability analysis of multiple KGEMs across several datasets. We find that high-performance models actually produce divergent predictions at the triple level and highly variable embedding spaces. By isolating stochastic factors (i.e., initialization, triple ordering, negative sampling, dropout, hardware), we show that each independently induces instability of comparable magnitude. Furthermore, for a given model, hyperparameter configurations with better MRR are not guaranteed to be more stable. Moreover, voting, albeit a known remediation mechanism, only provides a limited enhancement of stability. These findings highlight critical limitations of current benchmarking protocols, and raise concerns about the reliability of KGEMs for knowledge graph completion.

Fichier principal
Vignette du fichier
_ESWC26__Link_Prediction_or_Perdition__the_Seeds_of_Instability_in_Knowledge_Graph_Embeddings.pdf (1.04 Mo) Télécharger le fichier

Dates et versions

hal-05542928 , version 1 (09-03-2026)
hal-05542928 , version 2 (10-03-2026)

Licence

Identifiants

  • HAL Id : hal-05542928 , version 2

Citer

Guillaume Méroué, Fabien Gandon, Pierre Monnin. Link Prediction or Perdition: the Seeds of Instability in Knowledge Graph Embeddings. ESWC 2026 - 23rd European Semantic Web Conference, May 2026, Dubrovnik, Croatia. ⟨hal-05542928v2⟩
1554 Consultations
294 Téléchargements

Partager

  • More