AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Soft Label Pruning and Quantization for Large-Scale Dataset Distillation

Lingao Xiao, Yang He

TPAMI 2026 · Label-efficient and robust learning › Knowledge distillation · approach: Patient consistent distillation · 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
add class-wise matching (ablation)SRe2L baseline · ImageNet-1K IPC50 (ResNet-18), 50x labelstop-1 accuracy (%): 34.8 -> 37.7Improved
add class-wise BN supervision (ablation)SRe2L + class-wise matching · ImageNet-1K IPC50 (ResNet-18), 50x labelstop-1 accuracy (%): 37.7 -> 39.7Improved
add batch-level label pruning (ablation)LPQLD pipeline · ImageNet-1K IPC50 (ResNet-18), 50x labelstop-1 accuracy (%): 39.7 -> 43.1Improved
add dynamic knowledge reuse (temperature annealing) (ablation)LPQLD pipeline · ImageNet-1K IPC50 (ResNet-18), 50x labelstop-1 accuracy (%): 43.1 -> 47.8Improved
add label quantization (top-k logits) (ablation)LPQLD pipeline · ImageNet-1K IPC50 (ResNet-18), 50x labelstop-1 accuracy (%): 47.8 -> 49.7Improved
add calibrated student-teacher alignment (ablation)LPQLD pipeline · ImageNet-1K IPC50 (ResNet-18), 50x labelstop-1 accuracy (%): 49.7 -> 51.3Improved
add class-wise matching alone at 100x labels (ablation)SRe2L baseline · ImageNet-1K IPC50, 100x labelstop-1 accuracy (%): 25.4 -> 22.6Hurt
use class-wise batching and BN supervision for synthesisSRe2L / CDA images · ImageNet-1K, cosine similarity (IPC50)feature cosine similarity (lower better): 0.841 -> 0.796Improved
add 10x label pruning plus 8x quantization instead of pruning only at ~80xLPLD vs LPQLD · ImageNet-1K IPC10top-1 accuracy (%): 20.1 -> 27.3Improved
use quantization after 10x pruning rather than 40x pruning onlyLPLD · ImageNet-1K IPC50top-1 accuracy (%): 46.7 -> 51.3Improved
reduce top-k from 250 to 10 with 10x pruning (ablation)LPQLD · ImageNet-1K IPC10top-1 accuracy (%): 32.2 -> 20.0Hurt
replace FKD Marginal Smoothing top-5 quantization with LPQLD pruning+quantizationFKD quantization · ImageNet-1K IPC10top-1 accuracy (%): 7.5 -> 29.6Improved

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