Robust Fine-Tuning from Non-Robust Pretrained Models: Mitigating Suboptimal Transfer With Epsilon-Scheduling

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Ngnawé, Jonas, Heuillet, Maxime, Sahoo, Sabyasachi, Pequignot, Yann, Ahmad, Ola, Durand, Audrey, Precioso, Frédéric, Gagné, Christian
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866910051197779968
author Ngnawé, Jonas
Heuillet, Maxime
Sahoo, Sabyasachi
Pequignot, Yann
Ahmad, Ola
Durand, Audrey
Precioso, Frédéric
Gagné, Christian
author_facet Ngnawé, Jonas
Heuillet, Maxime
Sahoo, Sabyasachi
Pequignot, Yann
Ahmad, Ola
Durand, Audrey
Precioso, Frédéric
Gagné, Christian
contents Fine-tuning pretrained models is a standard and effective workflow in modern machine learning. However, robust fine-tuning (RFT), which aims to simultaneously achieve adaptation to a downstream task and robustness to adversarial examples, remains challenging. Despite the abundance of non-robust pretrained models in open-source repositories, their potential for RFT is less understood. We address this knowledge gap by systematically examining RFT from such non-robust models. Our experiments reveal that fine-tuning non-robust models with a robust objective, even under small perturbations, can lead to poor performance, a phenomenon that we dub suboptimal transfer. In challenging scenarios (eg, difficult tasks, high perturbation), the resulting performance can be so low that it may be considered a transfer failure. We find that fine-tuning using a robust objective impedes task adaptation at the beginning of training and eventually prevents optimal transfer. However, we propose a novel heuristic, Epsilon-Scheduling, a schedule over perturbation strength used during training that promotes optimal transfer. Additionally, we introduce expected robustness, a metric that captures performance across a range of perturbations, providing a more comprehensive evaluation of the accuracy-robustness trade-off for diverse models at test time. Extensive experiments on a wide range of configurations (six pretrained models and five datasets) show that Epsilon-Scheduling successfully prevents suboptimal transfer and consistently improves expected robustness.
format Preprint
id arxiv_https___arxiv_org_abs_2509_23325
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Robust Fine-Tuning from Non-Robust Pretrained Models: Mitigating Suboptimal Transfer With Epsilon-Scheduling
Ngnawé, Jonas
Heuillet, Maxime
Sahoo, Sabyasachi
Pequignot, Yann
Ahmad, Ola
Durand, Audrey
Precioso, Frédéric
Gagné, Christian
Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
Fine-tuning pretrained models is a standard and effective workflow in modern machine learning. However, robust fine-tuning (RFT), which aims to simultaneously achieve adaptation to a downstream task and robustness to adversarial examples, remains challenging. Despite the abundance of non-robust pretrained models in open-source repositories, their potential for RFT is less understood. We address this knowledge gap by systematically examining RFT from such non-robust models. Our experiments reveal that fine-tuning non-robust models with a robust objective, even under small perturbations, can lead to poor performance, a phenomenon that we dub suboptimal transfer. In challenging scenarios (eg, difficult tasks, high perturbation), the resulting performance can be so low that it may be considered a transfer failure. We find that fine-tuning using a robust objective impedes task adaptation at the beginning of training and eventually prevents optimal transfer. However, we propose a novel heuristic, Epsilon-Scheduling, a schedule over perturbation strength used during training that promotes optimal transfer. Additionally, we introduce expected robustness, a metric that captures performance across a range of perturbations, providing a more comprehensive evaluation of the accuracy-robustness trade-off for diverse models at test time. Extensive experiments on a wide range of configurations (six pretrained models and five datasets) show that Epsilon-Scheduling successfully prevents suboptimal transfer and consistently improves expected robustness.
title Robust Fine-Tuning from Non-Robust Pretrained Models: Mitigating Suboptimal Transfer With Epsilon-Scheduling
topic Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.23325