AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

PromptLoop: Plug-and-Play Prompt Refinement via Latent Feedback for Diffusion Model Alignment

Suhyeon Lee, Jong Chul Ye

CVPR 2026 · Reinforcement learning › RL from feedback · approach: RL fine-tuning of generative models

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
train the MLLM policy with GRPO instead of using it untrained (ablation)PromptLoop on SD1.5 · SD1.5 ImageRewardImageReward: -0.2315 -> 0.4344Improved
use an untrained MLLM to refine prompts (ablation)SD1.5 · SD1.5 ImageRewardImageReward: 0.0816 -> -0.2315Hurt
add multiple prompt refinements (5 steps) within one diffusion trajectory (ablation)PromptLoop on SD1.5 · SD1.5 ImageRewardImageReward: 0.4344 -> 0.4912Improved
add training-time visual (latent) feedback (ablation)PromptLoop on SD1.5 · SD1.5 ImageRewardImageReward: 0.4912 -> 0.6320Improved
increase the number of prompt refinement steps at inference (ablation)PromptLoop · SDXL and SD1.5 ImageRewardImageRewardImproved
more refinement steps when trained without visual feedback (ablation)PromptLoop · SDXL and SD1.5 ImageRewardImageRewardNo clear effect
add PromptLoop on top of Diffusion-DPO fine-tuned modelSDXL Diffusion-DPO · SDXL ImageRewardImageReward: 0.9921 -> 1.2898Improved
add PromptLoop on top of NPNetSDXL NPNet · SDXL ImageRewardImageReward: 0.7357 -> 1.1213Improved
apply policy trained on SD1.5 to unseen SDXLPromptLoop · SD1.5-trained, SDXL eval ImageRewardImageReward: 0.7244 -> 1.0859Improved
replace PromptLoop by GPT-5 or Qwen2.5-VL-3B prompt refinement on SD1.5SD1.5 · SD1.5 ImageRewardImageReward: 0.0816 -> -0.0950 (GPT-5), -0.1720 (Qwen2.5-VL-3B)Hurt
PromptLoop vs RePrompt under composite rewardSDXL-turbo · SDXL-turbo composite rewardImageReward: 0.7876 -> 0.8516Improved
use Qwen2.5-VL-3B prompt refinement without RL on SDXL-turboSDXL-turbo · SDXL-turbo composite rewardImageReward: 0.7769 -> 0.6417Hurt

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