Na minha lista:
Detalhes bibliográficos
Autor principal: Morrison, Kemar Armando
Formato: Recurso digital
Idioma:
Publicado em: Zenodo 2026
Acesso em linha:https://doi.org/10.5281/zenodo.19058064
Tags: Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!
Sumário:
  • <p>This dataset and experiment showcase the behavior of a controlled large language model (LLM) under a governance-layer framework originally validated in a synthetic harness. While the synthetic system exhibited fully stable (“GREEN”) outputs even under extreme stress, scaling to a full LLM introduces observable variability in the form of RED spikes, AMBER/BLUE transitional states, and extended GREEN stability zones. These patterns reveal emergent instabilities, cumulative drift, and dynamic recovery points, offering an empirical view of governance-layer performance at scale. The visualization provided highlights these zones, making the underlying stability dynamics immediately interpretable. This work demonstrates that RED spikes do not indicate failure but rather signify active governance response to high-dimensional LLM variability, validating the robustness and adaptive behavior of the system under realistic operating conditions.</p>