AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

HSIC Bottleneck for Cross-Generator and Domain-Incremental Synthetic Image Detection

Chin-Chia Yang, Yung-Yu Chuang, Hwann-Tzong Chen, Tyng-Luh Liu

ICLR 2026 · Image generation › Autoregressive and latent generation · approach: Generalizable fake image detection · 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
add intermediate ViT features (ablation)HSIC bottleneck detector · Trained on SDV1.4mACC (%): 89.20 -> 92.22Improved
add intermediate ViT features (ablation)HSIC bottleneck detector · Trained on ProGANmACC (%): 83.30 -> 89.22Improved
add HSIC(y,z) term (ablation)HSIC bottleneck with intermediates · Trained on SDV1.4mACC (%): 92.22 -> 93.39Improved
add HSIC(x,z) term on top of HSIC(y,z) and intermediates (ablation)HSIC bottleneck · Trained on SDV1.4mACC (%): 93.39 -> 93.86Improved
add HSIC(x,z) term on top of HSIC(y,z) and intermediates (ablation)HSIC bottleneck · Trained on ProGANmACC (%): 90.67 -> 91.07Improved
add HSIC(x,z) alone (ablation)CLIP detector · Trained on SDV1.4mACC (%): 89.20 -> 89.75No clear effect
use RBF median-bandwidth kernel instead of cosine (ablation)HSIC bottleneck · Trained on SDV1.4mACC (%): 88.70 -> 93.86Improved
use RBF median-bandwidth kernel instead of RBF sigma=1 (ablation)HSIC bottleneck · Trained on ProGANmACC (%): 88.82 -> 91.07Improved
add HSIC relevance to replay exemplar selection (ablation)HGR · SDV1.4 startmACC (%): 93.80 -> 94.38Improved
add HSIC relevance to replay exemplar selection (ablation)HGR · ProGAN startmACC (%): 86.23 -> 88.63Improved
use HGR replay instead of CBRSdomain-incremental detector · SDV1.4 startmACC (%): 92.66 -> 94.38Improved
use CLIP ViT-L/14 instead of DINOv2 ViT-L/14 (ablation)HSIC bottleneck · cross-generator detectionaccuracy: 10 to 30 percentage pointsImproved

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