![[2602.17826] Ontology-Guided Neuro-Symbolic Inference: Grounding Language Models with Mathematical Domain Knowledge](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.17826] Ontology-Guided Neuro-Symbolic Inference: Grounding Language Models with Mathematical Domain Knowledge
Summary
This article explores the integration of formal domain ontologies into language models to enhance their reliability in mathematical reasoning, addressing limitations like hallucination and brittleness.
Why It Matters
As language models become increasingly used in critical fields, ensuring their reliability is paramount. This research highlights the potential of neuro-symbolic approaches to improve model performance through structured knowledge, particularly in high-stakes environments where accuracy is crucial.
Key Takeaways
- Formal domain ontologies can enhance language model reliability.
- Neuro-symbolic pipelines can improve performance in specialized fields.
- Retrieval quality is critical; irrelevant context can degrade model output.
- The study uses mathematics as a proof of concept for the proposed method.
- Challenges remain in balancing context relevance and model performance.
Computer Science > Artificial Intelligence arXiv:2602.17826 (cs) [Submitted on 19 Feb 2026] Title:Ontology-Guided Neuro-Symbolic Inference: Grounding Language Models with Mathematical Domain Knowledge Authors:Marcelo Labre View a PDF of the paper titled Ontology-Guided Neuro-Symbolic Inference: Grounding Language Models with Mathematical Domain Knowledge, by Marcelo Labre View PDF HTML (experimental) Abstract:Language models exhibit fundamental limitations -- hallucination, brittleness, and lack of formal grounding -- that are particularly problematic in high-stakes specialist fields requiring verifiable reasoning. I investigate whether formal domain ontologies can enhance language model reliability through retrieval-augmented generation. Using mathematics as proof of concept, I implement a neuro-symbolic pipeline leveraging the OpenMath ontology with hybrid retrieval and cross-encoder reranking to inject relevant definitions into model prompts. Evaluation on the MATH benchmark with three open-source models reveals that ontology-guided context improves performance when retrieval quality is high, but irrelevant context actively degrades it -- highlighting both the promise and challenges of neuro-symbolic approaches. Comments: Subjects: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Symbolic Computation (cs.SC) Cite as: arXiv:2602.17826 [cs.AI] (or arXiv:2602.17826v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2602.17826 Focus to learn more arXiv...