Problem
Research questions and friction points this paper is trying to address.
e-commerce image generation
prompt quality
credit assignment
vision-language models
fine-grained optimization
Element-Aware Group Learning for E-Commerce Image Generation
🤖 AI Summary
This work addresses the limited fine-grained control in prompt optimization for e-commerce image generation by proposing the EAGLE-GRPO method, which decomposes holistic rewards into predefined design elements—such as composition, background, and selling-point presentation—to enable element-level credit assignment. Built upon the Group Relative Policy Optimization framework, the approach formulates element-wise advantage estimation as a kernel ridge regression problem and derives a closed-form solution, eliminating the need for additional rollouts or separate credit models while ensuring both computational efficiency and interpretability. Experimental results demonstrate that the generated prompts significantly enhance e-commerce image quality and continue to improve steadily in later training stages, effectively mitigating premature convergence plateaus.
📝 Abstract
Recent advances in image generation and editing have made prompt quality a key bottleneck for e-commerce creatives. Vision-language models (VLMs) can generate image-editing prompts from product images and metadata, but further improving their prompt-writing capabilities requires post-training with feedback from the generated images. Group Relative Policy Optimization (GRPO) is a natural framework for such outcome-level reward optimization. However, it assigns credit only at the full-prompt level, even though image quality often depends on specific design elements such as composition, background, and the presentation of selling points. Existing fine-grained credit assignment methods typically require step-level supervision or learned critics. To address this, we propose EAGLE-GRPO (Element-Aware Group Learning for E-Commerce Image Generation), which decomposes the group-centered reward over predefined elements. We cast element-level credit assignment as a kernel ridge regression problem and derive a closed-form solution, without additional rollouts or separate credit-assignment models. This yields interpretable per-element advantages and more precise policy updates. Experiments show that EAGLE-GRPO sustains performance gains over more training steps before plateauing and generates prompts that produce higher-quality e-commerce images than competitive VLM prompt-writing baselines.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Element-Aware Credit Assignment
Group Relative Policy Optimization
Vision-Language Models
Prompt Optimization
Kernel Ridge Regression
🔎 Similar Papers
No similar papers found.
