Ernesto Verdugo Independent Researcher, Verdugo Labs
This page catalogs peer-style working papers and technical research authored by Ernesto Verdugo on stability, recursion, and attractor dynamics in large-scale artificial intelligence systems.
Last updated: January, 2026

### Stable Attractor Dynamics in Large Language Models Under Controlled Perturbation ConditionsErnesto Verdugo (2025) Version: v1.0 Type: Technical Working Paper / Preprint

Abstract:This paper examines stable attractor dynamics emerging in large language models (LLMs) when exposed to structured, recursive perturbation protocols. Through multi-specimen comparative testing across state-of-the-art architectures, it identifies reproducible patterns of behavioral invariance under controlled recursion. The work introduces a falsifiable, architecture-neutral framework for analyzing stability, perturbation resilience, and collapse thresholds in high-parameter AI systems.
[Download PDF]