The Ontological Frontier: Large Language Model Ontological Reasoning

By Dr. Brent Allen Jensen

The authors glossed over the implication that their subgraph modelling path length, N=3 for arXiv, can result in smaller subgraphs which are easier to model accurately. This principle buried three levels deep in their data is a significant discovery. It suggests that the performance of OLLM on arXiv may be sensitive to the choice of the path length, and that fine-tuning this parameter may improve its accuracy. By carefully selecting the path length for each arXiv document, researchers can potentially obtain more accurate ontologies with fewer training samples. This finding has important implications for large-scale ontology construction and could open up new avenues for using large amounts of data to build ontologies.

REFERENCES:
[1] Web Source. "End-to-End Ontology Learning with Large Language Models[2501.09686] Towards Large Reasoning Models: A Survey of ...Large Language Models in Bio-Ontology Research: A ReviewExploring large language models for ontology learningAccelerating knowledge graph and ontology engineering with ...Toward large reasoning models: A survey of reinforced ...Journal Title Large Language Models for ©The Author(s) 2025 ...." arXiv:7e3bc8b7 (2026). https://arxiv.org/abs/2410.23584