multi-domain-learning

Proposes a conditional diffusion-guided knowledge transfer framework for multi-domain knowledge graph completion, generating domain-general entity embeddings without suppressing domain-specific information, achieving 4.3% average MRR improvement over state-of-the-art methods.