BambooKG: A Neurobiologically-inspired Frequency-Weight Knowledge Graph

TL;DR

BambooKG is a neurobiologically-inspired knowledge graph with frequency-weighted edges that enhances multi-hop reasoning in retrieval-augmented language models by reducing information loss and capturing complex relationships.

Contribution

It introduces a novel frequency-weighted knowledge graph that incorporates non-triplet edges, improving reasoning capabilities over traditional triplet-based graphs.

Findings

Outperforms existing knowledge graphs in reasoning tasks

Reduces information loss in knowledge representation

Enhances multi-hop reasoning accuracy

Abstract

Retrieval-Augmented Generation allows LLMs to access external knowledge, reducing hallucinations and ageing-data issues. However, it treats retrieved chunks independently and struggles with multi-hop or relational reasoning, especially across documents. Knowledge graphs enhance this by capturing the relationships between entities using triplets, enabling structured, multi-chunk reasoning. However, these tend to miss information that fails to conform to the triplet structure. We introduce BambooKG, a knowledge graph with frequency-based weights on non-triplet edges which reflect link strength, drawing on the Hebbian principle of "fire together, wire together". This decreases information loss and results in improved performance on single- and multi-hop reasoning, outperforming the existing solutions.