analiza aiqbench
This research addresses the critical safety challenge of adversarial feature missingness in reinforcement learning, where adversaries deliberately occlude observations to degrade agent performance. By developing a novel adversarial training algorithm, the authors demonstrate improved robustness across standard MuJoCo benchmarks compared to baselines trained on random missing data. The work establishes a formal framework for distinguishing this specific threat from bounded perturbations and general missing data problems, offering a pathway toward more resilient real-world AI systems.
Admissable: Training Reinforcement Learning Agents against Adversarial Missingness
arXiv:2609.15297v1 Announce Type: new Abstract: In order to make Reinforcement Learning algorithms applicable in real world scenarios, safety must be ensured even under adverse operating conditions. In this work, we consider the challenge of adversarial feature missingness: a scenario in which an adversary occludes features from the agent's observation in order to reduce performance as much as possible. We formally define adversarial missingness for Reinforcement Learning and compare it to the related concepts of $\ell_\infty$-norm bounded adversarial perturbations and learning with missing data. We develop an adversarial training algorithm and show its effectiveness in increasing robustness against adversarial missingness on three MuJoCo benchmark environments. Compared to a baseline trained with random uniform missingness, our method achieves better robustness on all three tasks.
Abstract:In order to make Reinforcement Learning algorithms applicable in real world scenarios, safety must be ensured even under adverse operating conditions. In this work, we consider the challenge of adversarial feature missingness: a scenario in which an adversary occludes features from the agent's observation in order to reduce performance as much as possible. We formally define adversarial missingness for Reinforcement Learning and compare it to the related concepts of $\ell_\infty$-norm bounded adversarial perturbations and learning with missing data. We develop an adversarial training algorithm and show its effectiveness in increasing robustness against adversarial missingness on three MuJoCo benchmark environments. Compared to a baseline trained with random uniform missingness, our method achieves better robustness on all three tasks.
Machine Learning (cs.LG)
arXiv:2609.15297 [cs.LG]
(or arXiv:2609.15297v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.15297 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 211 cuvinte
Subiecte conexe
Inverting Self-Triggered Control: Adversarial Reinforcement Learning for Sparse Denial-of-Service Attacks
arXiv · 14 sept. 2026
The Continual Constitutional Convention: How language models revise the documents that govern them
AI Alignment Forum · 13 oct. 2026
Continual learning might make your blocking monitors nearly useless
AI Alignment Forum · 24 sept. 2026
What We're Up Against: An AI Safety Crash Course
LessWrong · 24 sept. 2026