AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

High-dimensional Analysis of Synthetic Data Selection

Parham Rezaei, Filip Kovačević, Francesco Locatello, Marco Mondelli

ICLR 2026 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Score-based generative modeling analysis · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
select synthetic data by covariance matching instead of DS3CIFAR-10 augmentation with StyleGAN2-Ada · CIFAR-10, ResNet-18 scratchaccuracy (%): 52.83 -> 54.00Improved
covariance matching vs random selection of synthetic dataCIFAR-10 StyleGAN2-Ada, distillation · CIFAR-10, ResNet-18 distillationaccuracy (%): 54.89 -> 59.77Improved
covariance matching vs best baseline with T2I synthetic dataCIFAR-10 T2I, distillation · CIFAR-10, ResNet-18 distillationaccuracy (%): 59.43 -> 59.17No clear effect
covariance matching vs DS3 on ImageNet-100 T2I imagesImageNet-100 ResNet-18 scratch · ImageNet-100, T2I modelsaccuracy (%): 53.51 -> 53.07No clear effect
augment with covariance-matched synthetic data vs no synthetic dataRxRx1 linear probe · RxRx1accuracy (%): 86.83 -> 90.00Improved
mean shift between synthetic and target data (theory, linear regression) (ablation)min-norm least squares with real+synthetic data · high-dimensional linear regressionexcess riskNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy