Invariant-Driven Structural Learning for Language-Modeling
Abstract
Exploration of an alternative approach to language modeling where learning is guided by structuralinvariants instead of minimizing loss in fixed architectures. We introduce a self-organizingprobabilistic automaton that can dynamically change its internal structure, including states,responsibilities, and routing, when constraints are violated.
Fields of Study & Subjects
Machine Learning and Algorithms
DNA and Biological Computing
Natural Language Processing Techniques
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 5 January 2026
- Language
- EN
- Document type
- Preprint