aiqbench analysis
This study establishes the first dedicated benchmark for machine unlearning within Vision Transformers (ViT), addressing a critical gap where prior research focused almost exclusively on CNNs. By evaluating diverse architectures like Swin-T and DINOv2 against various datasets, it reveals how transformer-specific memorization patterns impact the ability to safely remove biased or sensitive training data without degrading model performance. This foundational work enables reproducible comparisons of unlearning algorithms, which is essential for deploying safe and fair computer vision systems in real-world applications.
Benchmarking Unlearning for Vision Transformers
arXiv:2602.20114v2 Announce Type: replace-cross Abstract: Machine unlearning (MU) refers to the post-training capability to remove (the influence of) training examples that are incorrect, biased, or leak sensitive/private information. MU is now widely regarded as critical for building safe and fair AI. In parallel, research into transformer architectures for computer vision has been highly successful: Vision Transformers (VTs) increasingly emerge as strong alternatives to CNNs. Yet, MU research for vision tasks has largely centered on CNNs, not VTs. While MU benchmarks have been developed for LLMs, diffusion models, and CNNs, none currently exist for VTs. This work is the first to attempt this, benchmarking MU algorithm performance across different VT families (ViT, Swin-T, and DINOv2) and at different capacities. The work employs (i) different datasets, selected to assess the impacts of dataset scale and complexity; (ii) different MU algorithms, selected to represent fundamentally different approaches for MU; and (iii) both single-shot and continual unlearning protocols. Additionally, it focuses on benchmarking MU algorithms that leverage training data memorization, since leveraging memorization has been recently discovered to significantly improve the performance of previously SOTA algorithms. En route, the work characterizes how VTs memorize training data relative to CNNs, and assesses the impact of different memorization proxies on performance. The benchmark uses unified evaluation metrics that capture two complementary notions of forget quality along with accuracy on unseen (test) data and on retained data. Overall, this work offers a benchmarking basis, enabling reproducible, fair, and comprehensive comparisons of existing (and future) MU algorithms on VTs. Importantly, for the first time, it sheds light on how well existing algorithms work in VT settings, establishing a promising reference performance baseline.
Abstract:Machine unlearning (MU) refers to the post-training capability to remove (the influence of) training examples that are incorrect, biased, or leak sensitive/private information. MU is now widely regarded as critical for building safe and fair AI. In parallel, research into transformer architectures for computer vision has been highly successful: Vision Transformers (VTs) increasingly emerge as strong alternatives to CNNs. Yet, MU research for vision tasks has largely centered on CNNs, not VTs. While MU benchmarks have been developed for LLMs, diffusion models, and CNNs, none currently exist for VTs. This work is the first to attempt this, benchmarking MU algorithm performance across different VT families (ViT, Swin-T, and DINOv2) and at different capacities. The work employs (i) different datasets, selected to assess the impacts of dataset scale and complexity; (ii) different MU algorithms, selected to represent fundamentally different approaches for MU; and (iii) both single-shot and continual unlearning protocols. Additionally, it focuses on benchmarking MU algorithms that leverage training data memorization, since leveraging memorization has been recently discovered to significantly improve the performance of previously SOTA algorithms. En route, the work characterizes how VTs memorize training data relative to CNNs, and assesses the impact of different memorization proxies on performance. The benchmark uses unified evaluation metrics that capture two complementary notions of forget quality along with accuracy on unseen (test) data and on retained data. Overall, this work offers a benchmarking basis, enabling reproducible, fair, and comprehensive comparisons of existing (and future) MU algorithms on VTs. Importantly, for the first time, it sheds light on how well existing algorithms work in VT settings, establishing a promising reference performance baseline.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
arXiv:2602.20114 [cs.CV]
(or arXiv:2602.20114v2 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2602.20114 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 · 367 words
Related coverage
BanglaRhet: Benchmarking Classical and Transformer Models for Rhetorical and Persuasion Detection in Bangla Political Speech
arXiv · Oct 8, 2026
David vs. Goliath in Next Activity Prediction: Argmax vs. LSTM, Transformer, and LLM
arXiv · Jun 16, 2026
StabilityBench: Benchmarking Instability in LLMs
arXiv · Jul 24, 2026
ONE CYLinder: A Benchmark for Graph-Based Surrogate Modeling of Unsteady Bluff-Body Flows
arXiv · Sep 9, 2026