Boosting Robust AIGI Detection with Lora-based Pairwise Training

Ruiyang Xia, Qi Zhang, Yaowen Xu, Zhaofan Zou, Hao Sun, Zhongjiang He, Xuelong Li · 2026

Read Boosting Robust AIGI Detection with Lora-based Pairwise Training by Ruiyang Xia, Qi Zhang, Yaowen Xu, Zhaofan Zou, Hao Sun, Zhongjiang He, Xuelong Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-12307.txt.

Description

Xia and six colleagues propose LoRA-based pairwise training to keep AI-generated image detectors robust under severe real-world distortions. Their plan fine-tunes a visual foundation model, mimics distortion and size shifts to match test data, and splits generalization from robustness tuning through pairwise training. The approach placed in the NTIRE robust detection challenge.

Catalog record

CREATORRuiyang Xia, Qi Zhang, Yaowen Xu, Zhaofan Zou, Hao Sun, Zhongjiang He, Xuelong Li
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE33404
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.12307

Browse the full library · About the Tower of Babel