Governed Knowledge Bundles: Verified Persistent Memory for Stateless Language Models

Read full text Summarise with VIRAT
Authors: Andres Cardozo Badillo
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

The architecture was iteratively developed across two distinct language-model platforms over four months, progressing through 30+ bundle versions. Evaluated over 8+ sessions spanning 21 days on the final platform, GKB achieves 95–100% terminology fidelity, 96.4% state recall accuracy, and 0.971 identity continuity index across cold starts—addressing the semantic decay, attention drift, and identity fragmentation identified as open problems in recent surveys [Smith, 2026, Du, 2026]. Cross-platform recovery was validated with early failures in ungoverned transfers motivating the integrity verification and guided recovery protocols presented here. The architecture presented here provides a foundation for applications requiring not just recall, but verified, consistent, and recoverable operational state.

Fields of Study & Subjects

Topic Modeling Natural Language Processing Techniques Speech and dialogue systems

Publication details

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