AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Multimodal Dataset Distillation Made Simple by Prototype-guided Data Synthesis

Junhyeok Choi, Sangwoo Mo, Minwoo Chae

ICLR 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Recaptioning and text-supervised pretraining · 0 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
use PDS instead of optimization-based distillationTESLA-VL/LoRS · Flickr30K 300 pairs, ResNetIR@1: 10.3 -> 14.4Improved
synthesize images with unCLIP decoder from image prototype instead of retrieving real image (ablation)PDS · Flickr30K 100 pairs, ResNetIR@10: 28.7 -> 37.3Improved
use unCLIP decoder instead of CLIP-inversion pixel optimization (ablation)PDS · Flickr30K 100 pairs, ResNetIR@1: 4.4 -> 7.9Improved
use unCLIP decoder instead of CLIP inversion for generation time (ablation)PDS · Flickr30K 100 pairsgeneration time per image (s): 1477.71 -> 9.71Improved
filter embeddings to shared pairs when building prototypes (ablation)PDS · Flickr30K 100 pairs, ResNetIR@10: 28.1 -> 37.3Improved
add retrieved-caption condition to image prototype (ablation)PDS · Flickr30K 100 pairs, ResNetIR@10: 28.3 -> 37.3Improved
discard pairless clusters at larger distilled sizes (ablation)PDS · Flickr30K 1000 pairs, ResNetIR@10: 56.4 -> 58.9Improved
discard pairless clusters at 100 pairs (ablation)PDS · Flickr30K 100 pairs, ResNetIR@10: 37.3 -> 37.5No clear effect
use agglomerative clustering instead of mini-batch k-means (ablation)PDS · Flickr30K 100 pairs, ResNetIR@10: 37.3 -> 41.4Improved
replace BERT with CLIP text encoder (ablation)LoRS baseline · Flickr30K 100 pairs, ResNetIR@1: 2.0 -> 6.3Improved
use VAE encoder instead of CLIP encoder (D4M/MGD3 extended) (ablation)learning-free distillation · Flickr30K 100 pairs, ResNetIR@10: 37.3 -> 17.2 (MGD3)Hurt
vary guidance scale and sampling steps of decoder (ablation)PDS · Flickr30Kaggregated retrieval accuracyNo 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