AI Paper Atlas

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

You Don’t Need All That Attention: Surgical Memorization Mitigation in Text-to-Image Diffusion Models

Kairan Zhao, Eleni Triantafillou, Peter Triantafillou

ICML 2026 · Image generation › Text-to-image diffusion · approach: Compositional text-to-image guidance · 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
attenuate cross-attention at all detected spike tokens instead of only EOT/paddingCA attenuation vs Ren et al. · SD v2.0 template memorizationSSCD: 0.36 -> 0.19Improved
attenuate cross-attention at all detected spike tokens instead of only EOT/paddingCA attenuation vs Ren et al. · SD v1.4 template memorizationSSCD: 0.60 -> 0.53Improved
use CA attenuation alone (no GUARD repulsion term) (ablation)CA attenuation · SD v2.0 template memorizationFID: 245.850 vs 212.727 for CA-in-GUARDHurt
embed CA attenuation as the positive target of GUARD (attraction-repulsion guidance) (ablation)CA attenuation · SD v1.4 verbatim memorizationSSCD: 0.109 -> 0.079Improved
embed CA attenuation as the positive target of GUARD (ablation)CA attenuation · SD v1.4 verbatim memorizationCLIP score: 0.282 -> 0.266Hurt
embed CA attenuation as the positive target of GUARD (ablation)CA attenuation · SD v1.4 verbatim memorizationFID: 164.660 -> 158.115Improved
embed CA attenuation as the positive target of GUARD (ablation)CA attenuation · SD v2.0 template memorizationSSCD: 0.193 -> 0.193No clear effect
attenuate CA only at EOT token (prior approach)Ren et al. · SD v1.4 template memorizationSSCD: 0.602 vs 0.530 for CA attenuationHurt
CA-in-GUARD instead of best prior methodCA-in-GUARD · SD v1.4 template memorizationSSCD: 0.479 -> 0.517 (Han et al. best prior is lower, 0.479)Improved
CA-in-GUARD instead of no mitigationStable Diffusion · SD v1.4 verbatim memorizationSSCD: 0.875 -> 0.079Improved
CA-in-GUARD instead of no mitigationStable Diffusion · SD v1.4 verbatim memorizationCLIP score: 0.346 -> 0.266Hurt

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