Conversational Control with Ontologies for Large Language Models: A Lightweight Framework for Constrained Generation
2026
Read Conversational Control with Ontologies for Large Language Models: A Lightweight Framework for Constrained Generation (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-04450.txt.
Description
Gendron, Guibon and d'Aquin propose a lightweight framework for controlled LLM generation using ontologies, modeling key conversational aspects as constraints. Large language model agents are black-box, with weak predictability and personalization. The team tests two aspects, English proficiency level and content polarity profile, fine-tuning seven open-weight conversational models with a hybrid procedure. Their method beats pre-trained baselines, even on smaller models, while staying model-agnostic, lightweight and interpretable. Reusable control plans extend to new domains and goals, lifting strategy alignment.
Catalog record
| YEAR | 2026 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37357 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.04450