Accelerating Large-Scale Dataset Distillation via Exploration-Exploitation Optimization

The high cost and accessibility problem associated with large datasets hinder the development of large-scale visual recognition systems. Dataset Distillation addresses these problems by synthesizing compact surrogate datasets for efficient training, storage, transfer, and privacy preservation. The existing state-of-the-art diffusion-based dataset distillation methods face three issues: lack of theoretical justification, poor efficiency in scaling to high data volumes, and failure in data-free scenarios. To address these issues, we establish a theoretical framework that justifies the use of diffusion models by proving the equivalence between dataset distillation and distribution matching, and reveals an inherent efficiency limit in the dataset distillation paradigm. We then propose a Dataset Concentration (DsCo) framework that uses a diffusion-based Noise-Optimization (NOpt) method to synthesize a small yet representative set of samples, and optionally augments the synthetic data via "Doping", which mixes selected samples from the original dataset with the synthetic samples to overcome the efficiency limit of dataset distillation. DsCo is applicable in both data-accessible and data-free scenarios, achieving SOTA performances for low data volumes, and it extends well to high data volumes, where it nearly reduces the dataset size by half with no performance degradation.

Dataset distillation (DD) compresses a large training set into a small synthetic set, reducing storage and training cost, and has shown strong results on general benchmarks. Decoupled DD further improves efficiency by splitting the pipeline into pretraining, sample distillation, and soft-label generation. However, existing decoupled methods largely rely on coarse class-label supervision and optimize samples within each class in a nearly identical manner. On fine-grained datasets, this often yields distilled samples that (i) retain large intra-class variation with subtle inter-class differences and (ii) become overly similar within the same class, limiting localized discriminative cues and hurting recognition. To solve the above-mentioned problems, we propose FD$^{2}$, a dedicated framework for Fine-grained Dataset Distillation. FD$^{2}$ localizes discriminative regions and constructs fine-grained representations for distillation. During pretraining, counterfactual attention learning aggregates discriminative representations to update class prototypes. During distillation, a fine-grained characteristic constraint aligns each sample with its class prototype while repelling others, and a similarity constraint diversifies attention across same-class samples. Experiments on multiple fine-grained and general datasets show that FD$^{2}$ integrates seamlessly with decoupled DD and improves performance in most settings, indicating strong transferability. Code is available at https://github.com/Guang000/FD2.

Training machine learning models on massive datasets is expensive and time-consuming. Dataset distillation addresses this by creating a small synthetic dataset that achieves the same performance as the full dataset. Recent methods use diffusion models to generate distilled data, either by promoting diversity or matching training gradients. However, existing approaches produce redundant training signals, where samples convey overlapping information. Empirically, disjoint subsets of distilled datasets capture 80-90% overlapping signals. This redundancy stems from optimizing visual diversity or average training dynamics without accounting for similarity across samples, leading to datasets where multiple samples share similar information rather than complementary knowledge. We propose learnability-driven dataset distillation, which constructs synthetic datasets incrementally through successive stages. Starting from a small set, we train a model and generate new samples guided by learnability scores that identify what the current model can learn from, creating an adaptive curriculum. We introduce Learnability-Guided Diffusion (LGD), which balances training utility for the current model with validity under a reference model to generate curriculum-aligned samples. Our approach reduces redundancy by 39.1%, promotes specialization across training stages, and achieves state-of-the-art results on ImageNet-1K (60.1%), ImageNette (87.2%), and ImageWoof (72.9%). Our code is available on our project page https://jachansantiago.github.io/learnability-guided-distillation/.