K-ON: Stacking Knowledge On the Head Layer of Large Language Model
This addresses a problem for NLP and knowledge graph integration, offering a novel method to improve entity-level predictions in LLMs.
The paper tackles the granularity mismatch between knowledge graphs and natural language in large language models by proposing K-ON, which integrates knowledge graph knowledge using multiple head layers for next k-step prediction, resulting in outperforming state-of-the-art methods that incorporate text and other modalities.
Recent advancements in large language models (LLMs) have significantly improved various natural language processing (NLP) tasks. Typically, LLMs are trained to predict the next token, aligning well with many NLP tasks. However, in knowledge graph (KG) scenarios, entities are the fundamental units and identifying an entity requires at least several tokens. This leads to a granularity mismatch between KGs and natural languages. To address this issue, we propose K-ON, which integrates KG knowledge into the LLM by employing multiple head layers for next k-step prediction. K-ON can not only generate entity-level results in one step, but also enables contrastive loss against entities, which is the most powerful tool in KG representation learning. Experimental results show that K-ON outperforms state-of-the-art methods that incorporate text and even the other modalities.