XAGI: Persistent and Explainable Cognitive Architecture Beyond Episodic Language Models

This paper presents an architectural framework for artificial cognitive systems with

persistent state, internal regulation, and structural explainability, called XAGI (Extended

Explainable General Artificial Intelligence).

Unlike conventional language models that operate as reactive generation systems, the

proposed framework maintains cognitive continuity through explicit separation between

linguistic processing and internal state mechanisms. The architecture integrates

components of affective regulation, cumulative memory, and cognitive governance

operating on a persistent core.

The system was evaluated through prolonged operation tests under controlled conditions,

with experimental records showing stability of internal patterns and an absence of abrupt

variations in functional configuration within the analyzed time horizons.

The findings suggest the viability of persistent and auditable cognitive architectures as a

complement to the purely statistical paradigm in language models, providing initial

evidence for the study of cognitive systems with structural temporal continuity.

Specific implementation details, configuration parameters, and concrete algorithms

constitute intellectual property in the process of being protected and are not disclosed in

this document. Access to complete technical material is available under a confidentiality

agreement for collaborative research or evaluation purposes.