TRIX: A More Expressive Model for Zero-shot Domain Transfer in Knowledge Graphs

TL;DR

TRIX is a new fully inductive knowledge graph model that improves expressiveness and handles both entity and relation prediction in zero-shot settings, outperforming existing models and large language models in out-of-domain tasks.

Contribution

Introduces TRIX, a more expressive inductive model capable of zero-shot entity and relation prediction, advancing knowledge graph completion in unseen domains.

Findings

TRIX outperforms state-of-the-art fully inductive models in zero-shot tasks.

TRIX surpasses large-context LLMs in out-of-domain predictions.

TRIX provides strictly more expressive triplet embeddings.

Abstract

Fully inductive knowledge graph models can be trained on multiple domains and subsequently perform zero-shot knowledge graph completion (KGC) in new unseen domains. This is an important capability towards the goal of having foundation models for knowledge graphs. In this work, we introduce a more expressive and capable fully inductive model, dubbed TRIX, which not only yields strictly more expressive triplet embeddings (head entity, relation, tail entity) compared to state-of-the-art methods, but also introduces a new capability: directly handling both entity and relation prediction tasks in inductive settings. Empirically, we show that TRIX outperforms the state-of-the-art fully inductive models in zero-shot entity and relation predictions in new domains, and outperforms large-context LLMs in out-of-domain predictions. The source code is available at https://github.com/yuchengz99/TRIX.