AI Paper Atlas

Research map / Label-efficient and robust learning / OOD and anomaly detection

Anomaly-Related Residual Fields for Cross-domain Anomaly Detection

Kewei Gao, Jiayi Xie, Zhengda Shen, Weijun Qin, Lingxiang Jia, Kejia Chen, Zunlei Feng, Yijun Bei

CVPR 2026 · Label-efficient and robust learning › OOD and anomaly detection · approach: Unified multi-class anomaly detection

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 R (amplitude residual) to REF (ablation)REF+CFA · VisA (Macaroni1, Macaroni2, Pcb2 average)AUROC (%): 89.90 -> 99.48Improved
add M (directional offset) to REF (ablation)REF+CFA · VisA (Macaroni1, Macaroni2, Pcb2 average)AUROC (%): 86.66 -> 99.48Improved
add Q (path-integrated drift) to REF (ablation)REF+CFA · VisA (Macaroni1, Macaroni2, Pcb2 average)AUROC (%): 87.06 -> 99.48Improved
use REF features instead of raw diffusion residuals (ablation)REF+CFA · VisA (Macaroni1, Macaroni2, Pcb2 average)AUROC (%): 84.11 -> 99.48Improved
add CFA temporal alignment (ablation)REF+CFA · VisA (Macaroni1, Macaroni2, Pcb2 average)AUROC (%): 95.89 -> 99.48Improved
add CFA directional subspace alignment (ablation)REF+CFA · VisA (Macaroni1, Macaroni2, Pcb2 average)AUROC (%): 88.40 -> 99.48Improved
add CFA second-order feature alignment (ablation)REF+CFA · VisA (Macaroni1, Macaroni2, Pcb2 average)AUROC (%): 92.40 -> 99.48Improved
add the full CFA module (ablation)REF+CFA · VisA (Macaroni1, Macaroni2, Pcb2 average)AUROC (%): 81.43 -> 99.48Improved
REF+CFA vs best baseline across transfersREF+CFA · 9 MVTec/VisA/DAGM transfersAUROC (%): 82.21 -> 95.22 averageImproved

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