ReCDAP: Relation-Based Conditional Diffusion with Attention Pooling for Few-Shot Knowledge Graph Completion

TL;DR

ReCDAP tackles few-shot knowledge graph completion under long-tail relations by explicitly modeling positive and negative information through a relation-conditioned diffusion process. It combines a global GNN-based aggregator, a Bi-LSTM relation learner, a diffusion module conditioned on $c_{global}$ with FiLM, and an attention pooler to extract salient positive/negative features, producing a latent $z$ used in a TransE-style scoring. The method introduces an extended support set $\tilde{C}_r$ that includes a generated negative set and a label $r_l$ to separate distributions. Experiments on NELL and FB15K-237 in 5-shot settings show state-of-the-art performance and strong ablations confirm the necessity of both diffusion conditioning and attention pooling. The work notes computational overhead and memory inefficiency, pointing to future work on memory-efficient multi-hop aggregation and lighter models.

Abstract

Knowledge Graphs (KGs), composed of triples in the form of (head, relation, tail) and consisting of entities and relations, play a key role in information retrieval systems such as question answering, entity search, and recommendation. In real-world KGs, although many entities exist, the relations exhibit a long-tail distribution, which can hinder information retrieval performance. Previous few-shot knowledge graph completion studies focused exclusively on the positive triple information that exists in the graph or, when negative triples were incorporated, used them merely as a signal to indicate incorrect triples. To overcome this limitation, we propose Relation-Based Conditional Diffusion with Attention Pooling (ReCDAP). First, negative triples are generated by randomly replacing the tail entity in the support set. By conditionally incorporating positive information in the KG and non-existent negative information into the diffusion process, the model separately estimates the latent distributions for positive and negative relations. Moreover, including an attention pooler enables the model to leverage the differences between positive and negative cases explicitly. Experiments on two widely used datasets demonstrate that our method outperforms existing approaches, achieving state-of-the-art performance. The code is available at https://github.com/hou27/ReCDAP-FKGC.

Figures (1)

• Figure 1: Model overview