aiqbench analysis
This research demonstrates that adversarial training significantly boosts generalization and robustness in audio classification models facing distribution shifts, such as those caused by varying recording devices in bioacoustics. By testing ConvNeXt and AudioProtoPNet architectures on the BirdSet benchmark, the study confirms that output-space attacks yield up to a 10.5% relative improvement in mean cmAP while strengthening model stability. These findings offer practical insights for deploying reliable machine learning systems in noisy, real-world acoustic environments where data distribution changes are inevitable.
Adversarial Training Improves Generalization Under Distribution Shifts in Bird Sound Classification
arXiv:2507.13727v2 Announce Type: replace Abstract: Adversarial training is a promising strategy for enhancing robustness against adversarial attacks, but its impact on generalization under substantial distribution shifts in audio classification remains largely unexplored. We address this gap by investigating how adversarial training strategies improve generalization performance and adversarial robustness in audio classification. We study two architectures for multi-label bird sound classification: ConvNeXt, a strong Convolutional Neural Network (CNN) baseline, and AudioProtoPNet, a prototype-based model that has demonstrated state-of-the-art performance while providing inherent interpretability through learned prototypes. Experiments use BirdSet, a challenging benchmark for bird sound classification in bioacoustics. Bioacoustic recordings exhibit substantial covariate shift due to heterogeneous recording devices and acoustic environments. We compare adversarial training based on output-space attacks, which maximize classification loss, and embedding-space attacks, which maximize embedding dissimilarity. Both attack types are also used for robustness evaluation. Additionally, for AudioProtoPNet, the study assesses the stability of its learned prototypes under targeted embedding-space attacks. Results show that adversarial training, particularly using output-space attacks, improves clean test-data performance by up to 10.5% relative in mean cmAP and simultaneously strengthens the adversarial robustness of the models. These findings, although derived from the bird sound domain, suggest that adversarial training holds potential to enhance robustness against both strong distribution shifts and adversarial attacks in challenging audio classification settings.
Abstract:Adversarial training is a promising strategy for enhancing robustness against adversarial attacks, but its impact on generalization under substantial distribution shifts in audio classification remains largely unexplored. We address this gap by investigating how adversarial training strategies improve generalization performance and adversarial robustness in audio classification. We study two architectures for multi-label bird sound classification: ConvNeXt, a strong Convolutional Neural Network (CNN) baseline, and AudioProtoPNet, a prototype-based model that has demonstrated state-of-the-art performance while providing inherent interpretability through learned prototypes. Experiments use BirdSet, a challenging benchmark for bird sound classification in bioacoustics. Bioacoustic recordings exhibit substantial covariate shift due to heterogeneous recording devices and acoustic environments. We compare adversarial training based on output-space attacks, which maximize classification loss, and embedding-space attacks, which maximize embedding dissimilarity. Both attack types are also used for robustness evaluation. Additionally, for AudioProtoPNet, the study assesses the stability of its learned prototypes under targeted embedding-space attacks. Results show that adversarial training, particularly using output-space attacks, improves clean test-data performance by up to 10.5% relative in mean cmAP and simultaneously strengthens the adversarial robustness of the models. These findings, although derived from the bird sound domain, suggest that adversarial training holds potential to enhance robustness against both strong distribution shifts and adversarial attacks in challenging audio classification settings.
Machine Learning (cs.LG)
arXiv:2507.13727 [cs.LG]
(or arXiv:2507.13727v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2507.13727 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 · 303 words
Related coverage
Does Adversarial Training Improve Generalization in Multi-View VLAs? Revealing and Mitigating View Collapse
arXiv · Sep 29, 2026
Latent reasoning architectures would likely undermine CoT, our strongest oversight tool
AI Alignment Forum · Sep 23, 2026
Problems of Proliferation
LessWrong · Sep 29, 2026
[AINews] not much happened today
Latent Space · Oct 3, 2026