When Attributes Disagree: Gradient Conflict in Image Aesthetic Assessment

Ye Wang, Maocai Dai, Jiang Xie, Xiuli Bi, Fei Tao, Xiao Li, Hong Yu
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:125672-125697, 2026.

Abstract

Image Aesthetic Assessment (IAA) predicts an image’s overall aesthetic score, yet aesthetic is influenced by multiple attributes whose relative importance varies with image content and usage scenarios. Under end-to-end training with only overall-score supervision, attribute signals are blended, which can cause gradient conflict across samples dominated by different attributes, resulting in gradient cancellation and persistent systematic bias. To address these issues, we propose AGREE (Attribute-guided Gradient Routing for Establishing Agreement), which learns attribute-specific subspaces and performs gradient routing based on sample-wise attribute sensitivity estimated via perturbation analysis. AGREE further reduces feature coupling across attributes with semantic anchors and improves robustness via error-aware reweighting. Experiments on AVA, LAPIS, AADB, TAD66K, and PARA show consistent improvements over diverse IAA baseline models, and AGREE is plug-and-play for existing end-to-end IAA methods without modifying their original architectures. To our knowledge, this work is among the early efforts in IAA to systematically study gradient conflict and provide an effective solution. The code is available at https://dahat364.github.io/AGREE/.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-wang26aj, title = {When Attributes Disagree: Gradient Conflict in Image Aesthetic Assessment}, author = {Wang, Ye and Dai, Maocai and Xie, Jiang and Bi, Xiuli and Tao, Fei and Li, Xiao and Yu, Hong}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {125672--125697}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/wang26aj/wang26aj.pdf}, url = {https://proceedings.mlr.press/v306/wang26aj.html}, abstract = {Image Aesthetic Assessment (IAA) predicts an image’s overall aesthetic score, yet aesthetic is influenced by multiple attributes whose relative importance varies with image content and usage scenarios. Under end-to-end training with only overall-score supervision, attribute signals are blended, which can cause gradient conflict across samples dominated by different attributes, resulting in gradient cancellation and persistent systematic bias. To address these issues, we propose AGREE (Attribute-guided Gradient Routing for Establishing Agreement), which learns attribute-specific subspaces and performs gradient routing based on sample-wise attribute sensitivity estimated via perturbation analysis. AGREE further reduces feature coupling across attributes with semantic anchors and improves robustness via error-aware reweighting. Experiments on AVA, LAPIS, AADB, TAD66K, and PARA show consistent improvements over diverse IAA baseline models, and AGREE is plug-and-play for existing end-to-end IAA methods without modifying their original architectures. To our knowledge, this work is among the early efforts in IAA to systematically study gradient conflict and provide an effective solution. The code is available at https://dahat364.github.io/AGREE/.} }
Endnote
%0 Conference Paper %T When Attributes Disagree: Gradient Conflict in Image Aesthetic Assessment %A Ye Wang %A Maocai Dai %A Jiang Xie %A Xiuli Bi %A Fei Tao %A Xiao Li %A Hong Yu %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-wang26aj %I PMLR %P 125672--125697 %U https://proceedings.mlr.press/v306/wang26aj.html %V 306 %X Image Aesthetic Assessment (IAA) predicts an image’s overall aesthetic score, yet aesthetic is influenced by multiple attributes whose relative importance varies with image content and usage scenarios. Under end-to-end training with only overall-score supervision, attribute signals are blended, which can cause gradient conflict across samples dominated by different attributes, resulting in gradient cancellation and persistent systematic bias. To address these issues, we propose AGREE (Attribute-guided Gradient Routing for Establishing Agreement), which learns attribute-specific subspaces and performs gradient routing based on sample-wise attribute sensitivity estimated via perturbation analysis. AGREE further reduces feature coupling across attributes with semantic anchors and improves robustness via error-aware reweighting. Experiments on AVA, LAPIS, AADB, TAD66K, and PARA show consistent improvements over diverse IAA baseline models, and AGREE is plug-and-play for existing end-to-end IAA methods without modifying their original architectures. To our knowledge, this work is among the early efforts in IAA to systematically study gradient conflict and provide an effective solution. The code is available at https://dahat364.github.io/AGREE/.
APA
Wang, Y., Dai, M., Xie, J., Bi, X., Tao, F., Li, X. & Yu, H.. (2026). When Attributes Disagree: Gradient Conflict in Image Aesthetic Assessment. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:125672-125697 Available from https://proceedings.mlr.press/v306/wang26aj.html.

Related Material