Pith. sign in

REVIEW 7 cited by

DiM: Distilling Dataset into Generative Model

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2303.04707 v2 pith:J5UJ3XEI submitted 2023-03-08 cs.CV

classification cs.CV
keywords distillationdatasetgenerativeimageslargearchitecturesmodeltextbf
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
abstract

Dataset distillation reduces the network training cost by synthesizing small and informative datasets from large-scale ones. Despite the success of the recent dataset distillation algorithms, three drawbacks still limit their wider application: i). the synthetic images perform poorly on large architectures; ii). they need to be re-optimized when the distillation ratio changes; iii). the limited diversity restricts the performance when the distillation ratio is large. In this paper, we propose a novel distillation scheme to \textbf{D}istill information of large train sets \textbf{i}nto generative \textbf{M}odels, named DiM. Specifically, DiM learns to use a generative model to store the information of the target dataset. During the distillation phase, we minimize the differences in logits predicted by a models pool between real and generated images. At the deployment stage, the generative model synthesizes various training samples from random noises on the fly. Due to the simple yet effective designs, the trained DiM can be directly applied to different distillation ratios and large architectures without extra cost. We validate the proposed DiM across 4 datasets and achieve state-of-the-art results on all of them. To the best of our knowledge, we are the first to achieve higher accuracy on complex architectures than simple ones, such as 75.1\% with ResNet-18 and 72.6\% with ConvNet-3 on ten images per class of CIFAR-10. Besides, DiM outperforms previous methods with 10\% $\sim$ 22\% when images per class are 1 and 10 on the SVHN dataset.

Discussion (0). Sign in to comment.

Forward citations

Cited by 7 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Dataset Distillation Based on Saliency-Driven Prototype Alignment

    cs.CV 2026-07 conditional novelty 6.0 of 10

    Saliency-guided latent prototypes plus confidence-based hard-prototype refinement improve diffusion-based dataset distillation accuracy on ImageNet subsets, CIFAR, and ImageNet-1K without fine-tuning the generative backbone.

  2. Dynamic-Aware Video Distillation: Optimizing Temporal Resolution Based on Video Semantics

    cs.CV 2025-05 conditional novelty 6.0 of 10

    DAViD uses Q-learning to pick a per-class temporal resolution for synthetic videos in dataset distillation, improving distilled-video recognition accuracy on UCF101, HMDB51, SSv2, and K400.

  3. MGD$^3$: Mode-Guided Dataset Distillation using Diffusion Models

    cs.CV 2025-05 conditional novelty 6.0 of 10

    Guiding a pretrained diffusion model toward per-class K-means centroids during sampling improves distilled dataset accuracy on ImageNet subsets by up to 4.4% while avoiding diffusion fine-tuning.

  4. CONCORD: Concept-Informed Diffusion for Dataset Distillation

    cs.CV 2025-05 conditional novelty 6.0 of 10

    A training-free concept-informed diffusion method, using LLM-retrieved and CLIP-filtered visual descriptions, improves dataset distillation accuracy on ImageNet subsets and ImageNet-1K.

  5. Dataset Distillation by Influence Matching

    cs.CV 2026-07 reject novelty 5.0 of 10

    Inf-Match distills datasets by matching estimated parameter influence of real and synthetic data, reporting SOTA classification and retrieval, but with an unsupported theoretical core.

  6. Dataset Distillation via Vision-Language Category Prototype

    cs.CV 2025-06 conditional novelty 5.0 of 10

    A dataset distillation method that combines K-means image prototypes with LLM-generated text prototypes to synthesize small, high-accuracy training sets.

  7. Diversity-Driven Generative Dataset Distillation Based on Diffusion Model with Self-Adaptive Memory

    cs.LG 2025-05 conditional novelty 4.0 of 10

    Diversity-driven diffusion dataset distillation with self-adaptive memory eviction achieves SOTA accuracy on ImageWoof, ImageNette, and ImageIDC.

Pith tools