UniPercept: Towards Unified Perceptual-Level Image Understanding across Aesthetics, Quality, Structure, and Texture

Shuo Cao, Jiayang Li, Xiaohui Li, Yuandong Pu, Kaiwen Zhu, Yuanting Gao, Siqi Luo, Yi Xin, Qi Qin, Yu Zhou, Xiangyu Chen, Wenlong Zhang, Bin Fu, Yu Qiao, Yihao Liu
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:11272-11301, 2026.

Abstract

Multimodal large language models (MLLMs) have achieved remarkable progress in visual understanding tasks such as visual grounding, segmentation, and captioning. However, their ability to perceive perceptual-level image features remains limited. In this work, we present UniPercept-Bench, a unified framework for perceptual-level image understanding across three key domains: Aesthetics, Quality, Structure and Texture. We establish a hierarchical definition system and construct large-scale datasets to evaluate perceptual-level image understanding. Based on this foundation, we develop a strong baseline UniPercept trained via Domain-Adaptive Pre-Training and Task-Aligned RL, enabling robust generalization across both Visual Rating (VR) and Visual Question Answering (VQA) tasks. UniPercept outperforms existing MLLMs on perceptual-level image understanding and can serve as a plug-and-play reward model for text-to-image generation. This work defines perceptual-level image understanding in the era of MLLMs and, through the introduction of a comprehensive benchmark together with a strong baseline, provides a solid foundation for advancing perceptual-level multimodal image understanding.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-cao26i, title = {{U}ni{P}ercept: Towards Unified Perceptual-Level Image Understanding across Aesthetics, Quality, Structure, and Texture}, author = {Cao, Shuo and Li, Jiayang and Li, Xiaohui and Pu, Yuandong and Zhu, Kaiwen and Gao, Yuanting and Luo, Siqi and Xin, Yi and Qin, Qi and Zhou, Yu and Chen, Xiangyu and Zhang, Wenlong and Fu, Bin and Qiao, Yu and Liu, Yihao}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {11272--11301}, 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/cao26i/cao26i.pdf}, url = {https://proceedings.mlr.press/v306/cao26i.html}, abstract = {Multimodal large language models (MLLMs) have achieved remarkable progress in visual understanding tasks such as visual grounding, segmentation, and captioning. However, their ability to perceive perceptual-level image features remains limited. In this work, we present UniPercept-Bench, a unified framework for perceptual-level image understanding across three key domains: Aesthetics, Quality, Structure and Texture. We establish a hierarchical definition system and construct large-scale datasets to evaluate perceptual-level image understanding. Based on this foundation, we develop a strong baseline UniPercept trained via Domain-Adaptive Pre-Training and Task-Aligned RL, enabling robust generalization across both Visual Rating (VR) and Visual Question Answering (VQA) tasks. UniPercept outperforms existing MLLMs on perceptual-level image understanding and can serve as a plug-and-play reward model for text-to-image generation. This work defines perceptual-level image understanding in the era of MLLMs and, through the introduction of a comprehensive benchmark together with a strong baseline, provides a solid foundation for advancing perceptual-level multimodal image understanding.} }
Endnote
%0 Conference Paper %T UniPercept: Towards Unified Perceptual-Level Image Understanding across Aesthetics, Quality, Structure, and Texture %A Shuo Cao %A Jiayang Li %A Xiaohui Li %A Yuandong Pu %A Kaiwen Zhu %A Yuanting Gao %A Siqi Luo %A Yi Xin %A Qi Qin %A Yu Zhou %A Xiangyu Chen %A Wenlong Zhang %A Bin Fu %A Yu Qiao %A Yihao Liu %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-cao26i %I PMLR %P 11272--11301 %U https://proceedings.mlr.press/v306/cao26i.html %V 306 %X Multimodal large language models (MLLMs) have achieved remarkable progress in visual understanding tasks such as visual grounding, segmentation, and captioning. However, their ability to perceive perceptual-level image features remains limited. In this work, we present UniPercept-Bench, a unified framework for perceptual-level image understanding across three key domains: Aesthetics, Quality, Structure and Texture. We establish a hierarchical definition system and construct large-scale datasets to evaluate perceptual-level image understanding. Based on this foundation, we develop a strong baseline UniPercept trained via Domain-Adaptive Pre-Training and Task-Aligned RL, enabling robust generalization across both Visual Rating (VR) and Visual Question Answering (VQA) tasks. UniPercept outperforms existing MLLMs on perceptual-level image understanding and can serve as a plug-and-play reward model for text-to-image generation. This work defines perceptual-level image understanding in the era of MLLMs and, through the introduction of a comprehensive benchmark together with a strong baseline, provides a solid foundation for advancing perceptual-level multimodal image understanding.
APA
Cao, S., Li, J., Li, X., Pu, Y., Zhu, K., Gao, Y., Luo, S., Xin, Y., Qin, Q., Zhou, Y., Chen, X., Zhang, W., Fu, B., Qiao, Y. & Liu, Y.. (2026). UniPercept: Towards Unified Perceptual-Level Image Understanding across Aesthetics, Quality, Structure, and Texture. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:11272-11301 Available from https://proceedings.mlr.press/v306/cao26i.html.

Related Material