AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Anti-Backdoor Coreset Selection via Cumulative Entropy

Qi Zhao, Christian Wressnegger

ICML 2026 · Adversarial robustness and security › Backdoor and poisoning · approach: Detecting poisoned samples · 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
accumulate entropy over epochs (Cumulative Entropy) instead of epoch-wise entropy sampling (ablation)uncertainty-based coreset selection · CIFAR10 Blend, ResNet18poisoning rate of coresetImproved
use Cumulative Entropy rather than cumulative Margin or LeastConf (ablation)anti-backdoor coreset selection · CIFAR10 Blend, ResNet18ACC, ASR, poisoning rateImproved
add min-max normalization of per-epoch entropy in CENT (ablation)ABCS · CIFAR10 WaNet, selection ratio 0.4poisoning rate of coreset (%): 0.62 -> 0.50Improved
add warm-up phase before selection (ablation)CENT · CIFAR10 WaNetpoisoning rate of coresetImproved
add unlearning step with label smoothing (ablation)CENT + warm-up · CIFAR10 WaNetpoisoning rate of coresetImproved
use a very small label smoothing factor epsilon (ablation)ABCS · CIFAR10 WaNetASRHurt
vary label smoothing factor epsilon (ablation)ABCS · CIFAR10natural accuracyNo clear effect
exclude mis-predicted samples when computing the threshold (ablation)ABCS · Tiny-ImageNet Blendnatural accuracy (%): 38.53 -> 61.11Improved
use a short warm-up (Twa) (ablation)ABCS · CIFAR10ASRHurt
use a larger regularization gamma (>=0.1) (ablation)ABCS · CIFAR10ASRImproved
use epoch-wise EL2N instead of cumulative EL2N (ablation)EL2N coreset selection · CIFAR10 Blendpoisoning rate of coresetHurt
use ABCS on clean datasetsABCS vs naive training · CIFAR10 cleannatural accuracy (%): 94.64 -> 94.77No 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