AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

FARI: Robust One-Step Inversion for Watermarking in Diffusion Models

Jindong Yang, Han Fang, Weiming Zhang, Nenghai Yu, Kejiang Chen

ICLR 2026 · Image generation › Autoregressive and latent generation · approach: Robust diffusion watermarking · 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
use FARI one-step inversion instead of 50-step DDIM, adversarial distortionSD v1.5 Tree-Ring · Tree-Ring adversarialTPR@1e-3: 0.949 -> 0.997Improved
use FARI instead of one-step DDIMSD v1.5 Tree-Ring · Tree-Ring adversarialTPR@1e-3: 0.863 -> 0.997Improved
use FARI instead of 50-step DDIMSD v2.1 Gaussian Shading · Gaussian Shading adversarialbit accuracy: 0.9755 -> 0.9824Improved
use FARI instead of naive inversionSD v3.5M Tree-Ring · Tree-Ring adversarialTPR@1e-3: 0.161 -> 0.978Improved
use FARI instead of one-step DDIMSDXL-Turbo Tree-Ring · Tree-Ring adversarialTPR@1e-3: 0.201 -> 0.855Improved
use ExactDPM instead of 50-step DDIMSD v2.1 Tree-Ring · Tree-Ring adversarialTPR@1e-3: 0.962 -> 0.906Hurt
use BELM instead of 50-step DDIMSD v1.5 Tree-Ring · Tree-Ring adversarialTPR@1e-3: 0.949 -> 0.592Hurt
add adversarial fine-tuning to AMED-Solver (AMED-dagger) vs FARIAMED-Solver · SD v1.5 Gaussian Shading adversarialbit accuracy: 0.9656 -> 0.9834Hurt
train FARI with adversarial distortions (vs without) (ablation)FARI · Image reconstructionPSNR: 14.60 -> 10.61Hurt
increase LoRA rank above 1 (ablation)FARI · LoRA rank ablationperformanceNo clear effect
train with multi-step inversion (NFE>1) instead of one step (ablation)FARI · NFE ablationperformanceHurt
train with all noise types instead of none, testing unseen distortions (ablation)FARI · Noise-type generalizationMSE of reconstructed noiseImproved

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