AI Paper Atlas

Research map / Graph learning / GNN architectures

Is Graph Unlearning Ready for Practice? A Benchmark on Efficiency, Utility, and Forgetting

Samyak Jain, Ronak Kalvani, sainyam galhotra, Sayan Ranu

ICLR 2026 · Graph learning › GNN architectures · approach: Robustness, privacy, and fairness of GNNs

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 PROJECTOR customised linear GNN instead of GCN-based methodsunlearning on Cora · Cora (10% nodes)fidelity to GOLD: 0.99 vs MEGU 0.95Improved
evaluate with fidelity instead of accuracy alonePROJECTOR · Roman-Empire / Coraaccuracy vs fidelity: accuracy 0.84 vs GOLD 0.88; fidelity 0.99Mixed
use IDEA instead of MEGUunlearning · Cora (10% nodes)logit L2 distance: 3.05 -> 18.03Hurt
use MEGU instead of retraining from scratchunlearning · Ogbn-arxivtotal unlearning time (s): 9.90 -> 20.49Hurt
use GIF instead of retrainingunlearning · Ogbn-arxivtotal unlearning time (s): 9.90 -> 4.79Improved
use ETR instead of retrainingunlearning · Reddittotal unlearning time (s): 96.73 -> 54.86Improved
use COGNAC instead of retrainingunlearning · Reddittotal unlearning time (s): 96.73 -> 197.43Hurt
use GNNDELETE instead of GCN-based unlearnersunlearning · Corainversion attack AUROC: 0.78 vs MEGU 0.53Hurt
delete high-degree nodes instead of random nodesGNNDELETE · Corafidelity to GOLD: 0.79 -> 0.69Hurt
delete nodes of the 2nd most frequent class instead of randomGRAPHERASER · Corafidelity to GOLD: 0.88 -> 0.74Hurt
raise unlearn ratio to 0.8MEGU/GIF/IDEA · Corafidelity and logit distanceHurt

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