AI Paper Atlas

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

No Caption, No Problem: Caption-Free Membership Inference via Model-Fitted Embeddings

Joonsung Jeon, Woo Jae Kim, Suhyeon Ha, Sooel Son, Sung-eui Yoon

ICLR 2026 · Image generation › Text-to-image diffusion · approach: Inference-time T2I improvement · 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
replace ground-truth captions with VLM-generated captionsCLiD · MIA PokemonASR (%): 96.52 -> 77.55Hurt
condition on model-fitted embedding instead of VLM captionsMoFit vs CLiD · MIA PokemonASR (%): 77.55 -> 94.48Improved
condition on model-fitted embedding instead of ground-truth captionsMoFit vs CLiD (GT) · MIA MS-COCOASR (%): 86.50 -> 88.00Improved
MoFit versus CLiD with GT captionsMoFit · MIA PokemonASR (%): 96.52 -> 94.48Hurt
extract embedding from optimized surrogate x0+delta* instead of raw query image (ablation)MoFit · MIA MS-COCOTPR@1%FPR: 31.00 -> 47.00Improved
extract embedding from surrogate instead of random-noise-perturbed image (ablation)MoFit · MIA PokemonTPR@1%FPR: 10.34 -> 50.48Improved
use adversarial delta_MAX instead of surrogate optimization (ablation)MoFit · MIA PokemonASR (%): 75.63 -> 75.87No clear effect
fine-tune target with LoRA instead of full U-NetMIA attacks · MIA SD v1.4 Pokemon LoRAASR (%): 94.48 -> 58.50Hurt
fine-tune with Gaussian blur or JPEG augmentationMoFit · MIA MS-COCOASR (%): 88.00 -> 88.70 (blur) / 82.80 (JPEG)Mixed
use MoFit on SD v1.5 relative to VLM-captioned CLiDMoFit · MIA SD v1.5TPR@1%FPR: 4.18 -> 41.30Improved
use MoFit versus PFAMI on SD v1.5 AUCMoFit · MIA SD v1.5AUC: 78.15 -> 71.03Hurt

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