AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Curriculum Group Policy Optimization: Adaptive Sampling for Unleashing the Potential of Text-to-Image Generation

Baoteng Li, Xianghao Zang, Xinran Wang, Xiangyu Na, Zhixiang He, Hao Sun, Chi Zhang, Zhongjiang He, Tianwei Cao, Kongming Liang, Zhanyu Ma

CVPR 2026 · Image generation › Text-to-image diffusion · approach: T2I evaluation and controllable adapters · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add variance-based probability sampling (ablation)Flow-GRPO · GenEvaloverall (%): 94.42 -> 95.15Improved
add exploration balancing (ablation)CGPO · GenEvaloverall (%): 95.15 -> 95.74Improved
add category calibration (ablation)CGPO · GenEvaloverall (%): 95.74 -> 96.10Improved
replace Flow-GRPO uniform sampling with CGPO adaptive sampling (efficiency)Flow-GRPO · GenEvalGPU hours to 0.944: 160 GPU hours, twice the speedImproved
use reward variance instead of advantage magnitude as proxy (ablation)CGPO · GenEvaloverall score: 0.80 -> 0.83Improved
use reward variance instead of multi-criteria proxy (ablation)CGPO · GenEvaloverall score: 0.82 -> 0.83Improved
set calibration lambda to 10 instead of 5 (ablation)Category calibration · GenEvaloverall score: 0.95 -> 0.96Improved
set calibration lambda to 8 instead of 10 (ablation)Category calibration · GenEvaloverall score: 0.96 -> 0.94Hurt
set calibration lambda to 15 instead of 10 (ablation)Category calibration · GenEvaloverall score: 0.96 -> 0.94Hurt
replace Flow-GRPO with CGPOFlow-GRPO · GenEval overallscore: 0.94 -> 0.96Improved
replace Flow-GRPO with CGPOFlow-GRPO · GenEval attribute bindingscore: 0.82 -> 0.89Improved
replace Flow-GRPO with CGPOFlow-GRPO · T2I-CompBench++ Texturescore: 0.7298 -> 0.7521Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy