analiza aiqbench
This paper is primarily about AI: it proposes a method for choosing a smaller training set for diffusion models, using PCA in a pre-trained CLIP embedding space to pick representative images. The claimed gains are up to 14x less training data and up to 1607x faster subset selection, with "marginal" performance differences. Those numbers come from the authors' experiments, so readers should check the benchmarks, datasets, and baselines before treating them as general results.
EigDiff: Eigensample Diffusion - or how to significantly reduce training dataset size with marginal performance variation?
arXiv:2610.20999v1 Announce Type: cross Abstract: Diffusion models are notoriously expensive to train, requiring large-scale datasets to learn how to gradually reverse the noise from images over multiple small denoising steps. Current approaches focus on distilling the number of diffusion steps, neglecting potential redundancies of the training data. In this paper, we explore the potential of harnessing eigensamples to train diffusion models on significantly smaller datasets, while preserving model performance. To mitigate the sensitivity to affine transformations of eigenimages, we harness the latent space of a pre-trained CLIP model. We next compute latent eigenvectors by performing principal component analysis (PCA) in CLIP space. Further, we employ eigenvectors as reference points to guide the selection of a representative training set comprising comprising both exemplars (images whose embeddings are closest to eigenvectors) and diverse images. Our comprehensive experiments with different diffusion models applied on multiple datasets indicate that our approach (EigDiff) can reduce training data by up to 14x and accelerate subset selection by up to 1607x, with marginal performance differences.
Title:EigDiff: Eigensample Diffusion - or how to significantly reduce training dataset size with marginal performance variation?
Abstract:Diffusion models are notoriously expensive to train, requiring large-scale datasets to learn how to gradually reverse the noise from images over multiple small denoising steps. Current approaches focus on distilling the number of diffusion steps, neglecting potential redundancies of the training data. In this paper, we explore the potential of harnessing eigensamples to train diffusion models on significantly smaller datasets, while preserving model performance. To mitigate the sensitivity to affine transformations of eigenimages, we harness the latent space of a pre-trained CLIP model. We next compute latent eigenvectors by performing principal component analysis (PCA) in CLIP space. Further, we employ eigenvectors as reference points to guide the selection of a representative training set comprising comprising both exemplars (images whose embeddings are closest to eigenvectors) and diverse images. Our comprehensive experiments with different diffusion models applied on multiple datasets indicate that our approach (EigDiff) can reduce training data by up to 14x and accelerate subset selection by up to 1607x, with marginal performance differences.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2610.20999 [cs.CV]
(or arXiv:2610.20999v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2610.20999 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 · 282 cuvinte
Subiecte conexe
[AINews] Opus 5.5 is good at explainer videos
Latent Space · 29 sept. 2026
[AINews] The Future of Latent Space
Latent Space · 25 sept. 2026
Triton-Based Optimization of Video Sparse Attention on ROCm
AMD · 13 iul. 2026
Research Note: Filtering Subversion-Relevant Information From Pretraining Data Is Feasible
LessWrong · 5 oct. 2026