AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

MSAT-LDM: Toward Transferable High-Fidelity Watermarking for Latent Diffusion Model via Modular Self-Augmented Training

Lu Zhang, Liang Zeng

AAAI 2026 · Image generation › Autoregressive and latent generation · approach: Robust diffusion watermarking

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
train watermark module on free-generation (empty prompt) images instead of external LAION images (ablation)MSAT-LDM · few-shot transfer to GuoFengFID: 6.87 -> 4.88Improved
train watermark module on free-generation images instead of LAION images (ablation)MSAT-LDM · few-shot transfer to GuoFeng+ShuiMo LoRAFID: 13.2 -> 8.30Improved
train watermark module on free-generation images instead of LAION images (ablation)MSAT-LDM · few-shot transfer to RealisticPSNR: 37.2 -> 36.8No clear effect
transfer pretrained watermark module instead of re-initializing it before few-shot fine-tuning (ablation)MSAT-LDM · few-shot transfer to GuoFengFID: 150 -> 4.88Improved
transfer pretrained watermark module instead of re-initializing it before few-shot fine-tuning (ablation)MSAT-LDM · few-shot transfer to GuoFengadversarial bit accuracy: 0.99 -> 0.99No clear effect
MSAT-LDM vs FSW under same few-shot fine-tuningFSW baseline · few-shot transfer to GuoFengFID: 169 -> 4.88Improved
MSAT-LDM vs Stable Signature on few-shot transferStable Signature baseline · few-shot transfer to GuoFengFID: 11.7 -> 4.88Improved
MSAT-LDM vs prior diffusion-native watermarks for image qualitySD v1.5 · AI-generated promptsFID: 9.49 -> 2.35Improved
train on LDM-generated data (free generation) vs real LAION images (ablation)watermark module training · AI-generated test promptsFID and adversarial bit accuracyImproved
free generation vs LAION-prompt generated data as training distribution (ablation)watermark module training · AI-generated test promptsrobustness at small sample sizesMixed
training-free Gaussian Shading vs MSAT-LDM on geometric attackGaussian Shading · perspective warp attackbit accuracy: 0.67 vs 0.98Hurt

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