AI Paper Atlas

Research map / Continual learning / Continual fine-tuning of LLMs

Static and Sequential Malicious Attacks in the Context of Selective Forgetting

CHENXU ZHAO, Wei Qian, Zhitao Ying, Mengdi Huai

NeurIPS 2023 · Continual learning › Continual fine-tuning of LLMs · approach: Machine unlearning by selective forgetting · 4 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 the proposed optimization instead of random removal (RandSearch)static forgetting attack, first-order · Diabetes targetedattack success rate: 0.04 ± 0.03 -> 1.00 ± 0.00Improved
use the proposed optimization instead of random removal (RandSearch)static forgetting attack, first-order · CIFAR-10 targetedattack success rate: 0.08 ± 0.04 -> 0.80 ± 0.04Improved
use the proposed optimization instead of RandSearch against SISAstatic forgetting attack, SISA · Diabetes targetedattack success rate: 0.47 ± 0.12 -> 0.74 ± 0.07Improved
use the proposed optimization instead of RandSearch against SISAstatic forgetting attack, SISA · CIFAR-10 targetedattack success rate: 0.13 ± 0.05 -> 0.67 ± 0.08Improved
apply optimization framework when unlearned data is underrepresented (1% to 5% of CIFAR-10) (ablation)first-order and second-order unlearning · CIFAR-10attack success rateImproved
increase the percentage of unlearning samples from the target class (5% to 25%)static forgetting attack · Diabetesattack success rateImproved
attack a subpopulation of size 50 with unrolling SGDstatic forgetting attack · Diabetes targetedattack success rate: 1.0No clear effect
attack a subpopulation of size 50 with amnesiac unlearningstatic forgetting attack · CIFAR-10 targetedattack success rate: 0.85Hurt
choose subpopulation members by feature similarity instead of randomly (ablation)static forgetting attack · Diabetes and CIFAR-10 subpopulationsattack success rateImproved
unlearn points from less redundant clusters (10% sampled) instead of more redundant (50%) (ablation)static forgetting attack · Diabetes and CIFAR-10attack success rateImproved
use optimization over random selection of points within each cluster (ablation)static forgetting attack · Diabetes and CIFAR-10 diversity studyattack success rateImproved
apply the learned adversarial policy instead of random or no attacksequential forgetting attack, first-order · Synthetic, Adult, MNISTEuclidean distance to target modelImproved

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