[2603.00166] Exploring the AI Obedience: Why is Generating a Pure Color Image Harder than CyberPunk?

Computer Science > Computer Vision and Pattern Recognition arXiv:2603.00166 (cs) [Submitted on 26 Feb 2026] Title:Exploring the AI Obedience: Why is Generating a Pure Color Image Harder than CyberPunk? Authors:Hongyu Li, Kuan Liu, Yuan Chen, Juntao Hu, Huimin Lu, Guanjie Chen, Xue Liu, Guangming Lu, Hong Huang View a PDF of the paper titled Exploring the AI Obedience: Why is Generating a Pure Color Image Harder than CyberPunk?, by Hongyu Li and 8 other authors View PDF HTML (experimental) Abstract:Recent advances in generative AI have demonstrated remarkable ability to produce high-quality content. However, these models often exhibit "Paradox of Simplicity": while they can render intricate landscapes, they often fail at simple, deterministic tasks. To address this, we formalize Obedience as the ability to align with instructions and establish a hierarchical grading system ranging from basic semantic alignment to pixel-level systemic precision, which provides a unified paradigm for incorporating and categorizing existing literature. Then, we conduct case studies to identify common obedience gaps, revealing how generative priors often override logical constraints. To evaluate high-level obedience, we present VIOLIN (VIsual Obedience Level-4 EvaluatIoN), the first benchmark focused on pure color generation across six variants. Extensive experiments on SOTA models reveal fundamental obedience limitations and further exploratory insights. By establishing this framework, we aim ...