AI Paper Atlas

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

CSF: Black-box Fingerprinting via Compositional Semantics for Text-to-Image Models

Junhoo Lee, Mijin Koo, Nojun Kwak

CVPR 2026 · Image generation › Text-to-image diffusion · approach: T2I evaluation and controllable adapters · 2 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 compositional underspecified prompts instead of naive prompts (ablation)CSF · Name That Dataset human studybase model identification accuracy (%): 18 -> 71Improved
use compositional prompts instead of base prompts without compositional constraints (ablation)CSF · 13 fine-tuned variantsdominance test passImproved
use Wasserstein distance instead of JSD (ablation)CSF · 4 fine-tuned variantsattribution confidence: Flux-LoRA 77.3% -> 93.2%, Kandinsky-Naruto 43.2% -> 97.7%Improved
apply UCE animal concept removal to the suspect modelCSF · 9 animal probesposterior mean attribution: e.g. SD1.5-Animal-Erase 0.857 on SD1.5No clear effect
fine-tune with LoRA, DPO, merging, distillation or VAE swapCSF attribution · 13 fine-tuned variantsdominance test (CI low > 0.5): all 13 variants pass; lowest posterior mean 0.659 (SD1.5-DreamShaper)No clear effect

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