Delta-Survival: Structural Collapse and Metabolic Recovery in LLMs

Summarise with VIRAT
Authors: Akihito Sunagawa
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

Research project investigating structural collapse in Large Language Models caused by contradiction accumulation (delta), and a metabolic architecture (DeltaZero) that prevents this collapse through cognitive sleep — processing knowledge during idle time. Key findings: - Context rot is caused by contradiction accumulation, not context length (Exp35, 6 models, 4 vendors) - External metabolism prevents collapse: ON vs OFF +52.2pp, p

Fields of Study & Subjects

Topic Modeling Natural Language Processing Techniques Adversarial Robustness in Machine Learning

Publication details

Journal / source
OSF Preprints (OSF Preprints)
Publication date
29 March 2026
Language
EN
Document type
Preprint