AI Paper Atlas

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

STEDiff: Revealing the Spatial and Temporal Redundancy of Backdoor Attacks in Text-to-Image Diffusion Models

Yu Pan, Jiahao Chen, Lin Wang, Bingrong Dai, Wenjie Wang

ICLR 2026 · Image generation › Text-to-image diffusion · approach: Inference-time T2I improvement

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 STEBA (enrichment-guided weights and sensitive timesteps) instead of VillanDiffusionSD v1.5 · COCO-CaptionFID: 27.58 -> 22.06Improved
use STEBA instead of VillanDiffusionSD v1.5 · COCO-Captionattack success rate (%): 99.4 -> 99.6Improved
use STEBA instead of VillanDiffusionRealistic Vision v4.0 · COCO-Captionattack success rate (%): 98.7 -> 95.4Hurt
update only enrichment weights and few timesteps (STEBA)SD v1.5 · COCO-CaptionTRA / SRA: 15.07× TRA, 5.55× SRAImproved
restrict fine-tuning to STEBA mask across samplersSD v1.5 DDPM · attack memoryminimum memory (MB): 18250MB -> 4090MBImproved
use STEBA across six diffusion samplers (ablation)SD v1.5 · COCO-Captions 2017 valattack success rate: 97.60% -> 98.25% (DDPM)No clear effect
use STEDF instead of T2IShieldSD v1.5 and others · trigger vocabulariesBDR (%): 91.2 -> 98.8 (Words)Improved
use STEDF instead of T2IShield on random/garbled triggersSD v1.5 · Random/GarbledBDR (%): 88.7 -> 98.1Improved
train STEDF on SD v1.5 only and transfer to other models (ablation)STEDF · SD v2.1 and RV v4.0BDR (%): 99.6 -> 92.1 and 88.9Mixed
apply STEDF against the STEBA attack rather than baseline attacksSTEDF · STEBA triggersBDR (%): 99.8 -> 74.1 (Phrases)Hurt
use STEDF instead of T2IShield against STEBASTEDF · STEBA SymbolsBDR (%): 57.3 -> 83.2Improved

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