Trustworthy DD · Optimization and training recipes
ROME
ROME is Forged in Adversity: Robust Distilled Datasets via Information Bottleneck
Zheng Zhou
ICML 2025 · first public 2025-01-01
In one paragraph
Applies the Information Bottleneck principle to distillation robustness with two loss terms — a performance-aligned term to preserve clean accuracy and a robustness-aligned term that aligns feature distributions between synthetic and adversarially perturbed images — and introduces an Improved Robustness Ratio (I-RR) metric to evaluate DD robustness more precisely. Reports up to roughly 40% and 35% I-RR improvements over existing DD methods under white-box and black-box attacks respectively on CIFAR-10/100.
Where it sits
- Optimization and training recipes (Orthogonal design choices)
- Setting: Image classification
BibTeX (generated; prefer the venue's official entry)
@article{zhou2025rome,
title = {ROME is Forged in Adversity: Robust Distilled Datasets via Information Bottleneck},
author = {Zheng Zhou},
journal = {ICML 2025},
year = {2025}
}Nearby in Optimization and training recipes
PTM-ST — Multimodal Dataset Distillation via Phased Teacher Models
Shengbin Guo, Hang Zhao, Senqiao Yang et al. · ICLR 2026notableVision–languagepaper ↗code ↗