Method Detail: Aesbench

Back to Leaderboard
Benchmark: AesBench
Short name: Aesbench
Long name: An Expert Benchmark for Multimodal Large Language Models on Image Aesthetics Perception
Description: AesBench is an expert-level benchmark designed to comprehensively evaluate the aesthetic understanding and perception capabilities of Multimodal Large Language Models (MLLMs). It features the Expert-labeled Aesthetics Perception Database (EAPD), containing 2,800 diversely-sourced images (natural, artistic, and AI-generated) annotated by professional aesthetic experts. It measures model performance across four shallow-to-deep dimensions: AesP (Aesthetic Perception) AesE (Aesthetic Empathy) AesA (Aesthetic Assessment) AesI (Aesthetic Interpretation)
Reference: Yipo Huang, Quan Yuan, Xiangfei Sheng, Zhichao Yang, Haoning Wu, Pengfei Chen, Yuzhe Yang, Leida Li, Weisi Lin, AesBench: An Expert Benchmark for Multimodal Large Language Models on Image Aesthetics Perception. In arXiv preprint arXiv:2401.08276, 2024.
Last submitted: September 15, 2026
Published: September 15, 2026 at 06:49:35
Submissions: 3
Project page / code: https://github.com/yipoh/AesBench.git
Open source: No

Benchmark performance

Submission Date AesP (↑) AesE (↑) AesA1 (↑)
2026-09-15 07:06 0.0000 0.0000 0.0000
2026-09-15 07:01 0.0000 0.0000 0.0000
2026-09-15 06:56 0.0000 0.0000 0.0000