aiqbench analysis
This research proposes a geometry-aware deep learning framework designed to enhance adversarial robustness by optimizing the internal feature space of neural networks. By leveraging layer-wise local training, the approach aims to achieve manifold smoothness against various attack types while maintaining high accuracy through statistical mechanics models. The work offers significant theoretical insights into mitigating vulnerabilities in machine learning systems without relying on external hardware or specific consumer applications.
Geometric origin of adversarial vulnerability in deep learning
arXiv:2509.01235v2 Announce Type: replace Abstract: Balancing training accuracy and adversarial robustness has beeen a challenge since the birth of deep learning. Here, we introduce a geometry-aware deep learning framework that leverages layer-wise local training to sculpt the internal representations of deep neural networks. This framework promotes intra-class compactness and inter-class separation in feature space, leading to manifold smoothness and adversarial robustness against white or black box attacks. The performance can be explained by \blue{data-dependent statistical mechanics of integrating out the network parameters}, \blue{supplemented by a phenomenological model} with Hebbian coupling between elements of the hidden representation. Based on the current geometry-aware learning framework, the deep network can assimilate new information into existing knowledge structures while reducing representation interference.
Abstract:Balancing training accuracy and adversarial robustness has beeen a challenge since the birth of deep learning. Here, we introduce a geometry-aware deep learning framework that leverages layer-wise local training to sculpt the internal representations of deep neural networks. This framework promotes intra-class compactness and inter-class separation in feature space, leading to manifold smoothness and adversarial robustness against white or black box attacks. The performance can be explained by \blue{data-dependent statistical mechanics of integrating out the network parameters}, \blue{supplemented by a phenomenological model} with Hebbian coupling between elements of the hidden representation. Based on the current geometry-aware learning framework, the deep network can assimilate new information into existing knowledge structures while reducing representation interference.
Machine Learning (cs.LG); Statistical Mechanics (cond-mat.stat-mech); Neurons and Cognition (q-bio.NC)
arXiv:2509.01235 [cs.LG]
(or arXiv:2509.01235v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2509.01235 Focus to learn more arXiv-issued DOI via DataCite
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.
Full text under the publisher's open licence · 212 words
Related coverage
Resisting Adversarial Attacks in Deep Neural Networks using Diverse Decision Boundaries
arXiv · Oct 6, 2026
Projection sampling preserves capabilities but diminishes gains from alignment fine-tuning
LessWrong · Oct 16, 2026
Frontier Risk Report (February to March 2026)
METR · May 19, 2026
Frontier AI Risk Report (February–March 2026)
METR · May 19, 2026