Historical Genetic Logic as a Dynamical Coherence Judge for Large Language Models

Read full text Summarise with VIRAT
Authors: AKATERINH XENOPOULOU-TYROKOMOU, Epameinondas Xenopoulos
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Historical Genetic Logic as a Dynamical Coherence Judge for Large Language Models A Rigorous Formalization of Xenopoulos’ Dialectical Operators and Experimental Validation on LLM Self‑Contradiction Katerina XenopoulouIndependent Researcher, Kefalonia, GreeceORCID: 0009-0004-9057-7432Correspondence: katerinaxenopoulou@gmail.com Theoretical Foundation: Epameinondas Xenopoulos †Epistemology of Logic: Logic–Dialectic or Theory of Knowledge (2nd ed., 2024)ORCID: 0009-0000-1736-8555† In memoriam (1920–1994) DOI: 10.5281/zenodo.19263676 https://zenodo.org/uploads/19263676 ABSTRACT Internal self‑contradiction remains a critical failure mode in Large Language Models (LLMs), limiting their reliability in high‑stakes reasoning. While current mitigation strategies like Chain‑of‑Thought (CoT) prompting improve performance, they lack formal guarantees of logical stability. This paper introduces a novel framework for diagnosing and regulating LLM coherence by formalizing Historical Genetic Logic as a Nonlinear Dynamical System. We demonstrate that the reasoning process in autoregressive models can be modeled as a trajectory in a recursive metric space D=⋃n=0∞DnD=⋃n=0∞Dn with Dn+1=[0,1]2×Pfin(Dn)Dn+1=[0,1]2×Pfin(Dn). Our core theoretical contribution, the Xenopoulos Spectral Invariance Theorem (Theorem 7.1), proves that CoT prompting leaves the Lyapunov spectrum invariant, merely extending unstable trajectories without suppressing the underlying chaotic divergence. To address this, we propose the Xenopoulos Layer, a spectral feedback controller that dynamically intervenes in the Jacobian operator Fγ=F−γIFγ=F−γI. By enforcing a negative Lyapunov exponent λ1(γ)

Fields of Study & Subjects

Logic Reasoning and Knowledge Topic Modeling Natural Language Processing Techniques

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
26 March 2026
Language
EN
Document type
Preprint