Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ ZENODOarrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
ZENODO
Report . 2025
License: CC BY
Data sources: ZENODO
ZENODO
Report . 2025
License: CC BY
Data sources: Datacite
ZENODO
Report . 2025
License: CC BY
Data sources: Datacite
versions View all 2 versions
addClaim

"Prudential Gating Function v3: Multi-Seed Validation of a Risk-Aware Reward Shaping Mechanism for Reinforcement Learning"

Función de Puerta Prudencial v3: Validación Multi-Semilla de un Mecanismo de Recompensa Sensible al Riesgo para Aprendizaje por Refuerz
Authors: Rivera Garcia, Jose M;

"Prudential Gating Function v3: Multi-Seed Validation of a Risk-Aware Reward Shaping Mechanism for Reinforcement Learning"

Abstract

I present a multi-seed validation of the Prudential Gating Function (PGF) v3, a reward shaping mechanism designed to induce risk-aware behavior in reinforcement learning agents operating in stochastic environments. Across three independent random seeds in a 5×5 gridworld with moderate risk conditions (risk_scale=1.5), PGF v3 achieves a mean performance ratio of 38.93% ± 0.59% relative to a risk-blind control agent, with exceptional statistical reproducibility (coefficient of variation = 1.52%). This represents a cumulative +131.7% improvement over our initial baseline implementation Presento una validación multi-semilla de la Función de Puerta Prudencial (PGF) v3, un mecanismo de modelado de recompensas diseñado para inducir comportamiento sensible al riesgo en agentes de aprendizaje por refuerzo que operan en entornos estocásticos. A través de tres semillas aleatorias independientes en un mundo de rejilla (gridworld) de 5×5 con condiciones de riesgo moderado (risk_scale=1.5), PGF v3 logra una razón de desempeño promedio de 38.93% ± 0.59% en relación con un agente de control ciego al riesgo, con una reproducibilidad estadística excepcional (coeficiente de variación = 1.52%). Esto representa una mejora acumulada de +131.7% sobre nuestra implementación base inicial.

Keywords

reinforcement learning, Aprendizaje, multi-seed reproducibility, safe reinforcement learning, Risk Modeling, Modelo de Riesgo, Alineamiento, alignment tax, statistical validation, Artificial Intelligence, prudential behavior, AI Safety, risk-aware agents, Inteligencia Artificial, reward shaping

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Green
Related to Research communities