AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

MUNBa: Machine Unlearning via Nash Bargaining

Jing Wu, Mehrtash Harandi

ICCV 2025 · Image generation › Text-to-image diffusion · approach: Concept erasure in diffusion models · 10 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 MUNBa instead of SalUnunlearning · Celeb-HQ-307 (forget 10% identities)Avg. Gap: 0.60 -> 0.10Improved
use MUNBa instead of SalUnunlearning · CIFAR-10 (forget 10% samples)Avg. Gap: 1.24 -> 0.97Improved
use MUNBa instead of l1-sparseunlearning · SVHN (class-wise forgetting)Avg. Gap: 0.77 -> 1.40Hurt
use MUNBa instead of SHsunlearning · Celeb-HQ-307 (forget 50% identities)Avg. Gap: 0.51 -> 0.97Hurt
use MUNBa instead of SalUnunlearning · CIFAR-10 (forget 50% samples)Avg. Gap: 2.18 -> 5.18Hurt
use MUNBa instead of SHsCLIP unlearning · Oxford Pets (forget one class)Acc ImageNet: 37.97 -> 59.36Improved
use MUNBa instead of SalUnCLIP unlearning · Oxford Pets (forget one class)Acc ImageNet: 59.94 -> 59.36No clear effect
use MUNBa instead of FTCLIP unlearning · Oxford Pets (forget one class)Acc test (D_t): 91.14 -> 94.99Improved
use MUNBa instead of SalUnStable Diffusion unlearning · nudity forgettingFID: 25.06 -> 15.92Improved
use MUNBa instead of SalUnStable Diffusion unlearning · nudity forgettingattack success rate: 11.27 -> 3.52Improved
use MUNBa instead of SalUnStable Diffusion unlearning · Imagenette class-wise forgettingUA: 99.82 -> 99.94No clear effect
use MUNBa instead of FTunlearning · Celeb-HQ-307 (forget 10% identities)Acc forget (D_f): 99.94 -> 0.00Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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