Reducing AI Entropy: The Information Dynamics of Model Safety

Fuente: Zenodo
Guardado en:
Detalles Bibliográficos
Autor principal: Kugelmass, Joe
Formato: Recurso digital
Lenguaje:inglés
Publicado: Zenodo 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866901670047252480
author Kugelmass, Joe
author_facet Kugelmass, Joe
contents <p>Large language models (LLMs) exhibit a fundamental tension: fluent outputs coexist with unpredictable instabilities—hallucinations, semantic drift, and oscillatory responses. Existing safety mechanisms evaluate outputs post-hoc but do not monitor the underlying information dynamics during generation. We introduce ΔK, an information-theoretic metric measuring divergence and boundary flux across multiple stochastic samples of a model's output. Built atop this metric, the Synthesized Oracle implements a feedback control system for LLMs that identifies unstable segments, removes oscillatory components, and reconstructs coherent responses through entropy-aware optimization.</p> <p>Across 188 evaluations on a 3-billion parameter model (cogito:3b), the system demonstrates: (1) reliable instability detection through flux dynamics, (2) consistent entropy reduction via J-score minimization (53.5% of cases requiring optimization), and (3) successful reconstruction of approximately half (53.1%) of chaotic responses through surgical oscillator removal. These results suggest that model safety can be reframed from behavioral policing to dynamical regulation—treating entropy not as a defect to eliminate but as a signal to guide.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_17872999
institution Zenodo
language eng
publishDate 2025
publisher Zenodo
record_format zenodo
spellingShingle Reducing AI Entropy: The Information Dynamics of Model Safety
Kugelmass, Joe
instability detection
LLM alignment
generative AI
control theory
cybernetics
flux measurement
entropy
model safety
dynamical systems
large language models
artificial intelligence
LLMs
<p>Large language models (LLMs) exhibit a fundamental tension: fluent outputs coexist with unpredictable instabilities—hallucinations, semantic drift, and oscillatory responses. Existing safety mechanisms evaluate outputs post-hoc but do not monitor the underlying information dynamics during generation. We introduce ΔK, an information-theoretic metric measuring divergence and boundary flux across multiple stochastic samples of a model's output. Built atop this metric, the Synthesized Oracle implements a feedback control system for LLMs that identifies unstable segments, removes oscillatory components, and reconstructs coherent responses through entropy-aware optimization.</p> <p>Across 188 evaluations on a 3-billion parameter model (cogito:3b), the system demonstrates: (1) reliable instability detection through flux dynamics, (2) consistent entropy reduction via J-score minimization (53.5% of cases requiring optimization), and (3) successful reconstruction of approximately half (53.1%) of chaotic responses through surgical oscillator removal. These results suggest that model safety can be reframed from behavioral policing to dynamical regulation—treating entropy not as a defect to eliminate but as a signal to guide.</p>
title Reducing AI Entropy: The Information Dynamics of Model Safety
topic instability detection
LLM alignment
generative AI
control theory
cybernetics
flux measurement
entropy
model safety
dynamical systems
large language models
artificial intelligence
LLMs
url https://doi.org/10.5281/zenodo.17872999