AI Paper Atlas

Research map / Large language models / Alignment and preferences

BeautyGRPO: Aesthetic Alignment for Face Retouching via Dynamic Path Guidance and Fine-Grained Preference Modeling

Jiachen Yang, Xianhui Lin, Yi Dong, Zebiao Zheng, Xing Liu, Hong Gu, Yanmei Fang

CVPR 2026 · Large language models › Alignment and preferences · approach: Reward model design and preference models · 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 Dynamic Path Guidance to online RL (FlowGRPO) for retouchingFLUX.1 Kontext + LoRA · FFHQRNIQE: 15.024 -> 10.831Improved
add Dynamic Path Guidance to online RL (FlowGRPO) for retouchingFLUX.1 Kontext + LoRA · FFHQRMUSIQ: 4.271 -> 4.906Improved
add Dynamic Path Guidance to online RL (FlowGRPO) for retouchingFLUX.1 Kontext + LoRA · FFHQRArcFace: 0.882 -> 0.952Improved
apply BeautyGRPO RL on top of supervised LoRA fine-tuningFLUX.1 Kontext + LoRA · FFHQRNIMA: 4.694 -> 5.123Improved
apply BeautyGRPO RL on top of supervised LoRA fine-tuningFLUX.1 Kontext + LoRA · FFHQRFID: 1.735 -> 4.054Hurt
apply BeautyGRPO RL on top of supervised LoRA fine-tuningFLUX.1 Kontext + LoRA · FFHQRArcFace: 0.973 -> 0.952Hurt
use the fine-grained retouching reward model instead of EditReward (ablation)BeautyGRPO on FLUX.1 Kontext + LoRA · FFHQRNIMA: 4.754 -> 5.123Improved
use the fine-grained retouching reward model instead of EditScore (ablation)BeautyGRPO on FLUX.1 Kontext + LoRA · FFHQRTOPIQ: 0.621 -> 0.676Improved
use the fine-grained retouching reward model instead of UnifiedReward-Edit (ablation)BeautyGRPO on FLUX.1 Kontext + LoRA · FFHQRMUSIQ: 4.571 -> 4.906Improved
apply BeautyGRPO to Qwen-Image-Edit on top of LoRAQwen-Image-Edit · FFHQRMUSIQ: 4.745 -> 5.133Improved
apply BeautyGRPO to Qwen-Image-Edit on top of LoRAQwen-Image-Edit · FFHQRTOPIQ: 0.595 -> 0.664Improved
increase DPG sampling segments K from 3 to 5 (ablation)BeautyGRPO · FFHQRMUSIQ / NIMANo clear effect

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