AI Paper Atlas

Research map / Image generation / Editing and style transfer

Training-Free Reward-Guided Image Editing via Trajectory Optimal Control

Jinho Chang, Jaemin Kim, Jong Chul Ye

ICLR 2026 · Image generation › Editing and style transfer · approach: Inversion-based diffusion editing

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
optimize entire trajectory via adjoint guidance instead of inversion + posterior-mean guidance (TFG)Ours · Human preference, SD1.5ImageReward / LPIPS: ImageReward 1.7053 -> 1.8914; LPIPS 0.2927 -> 0.1717Improved
use trajectory control instead of gradient ascentOurs · Human preference, SD1.5HPSv2: 0.2247 -> 0.2526Improved
compare against gradient ascent on target rewardOurs · Human preference, SD1.5ImageReward: 1.9088 -> 1.8914Hurt
trajectory control vs FreeDoM in style transferOurs · Style transfer, SD1.5DINO style similarity: 0.2250 -> 0.2467Improved
trajectory control vs gradient ascent in counterfactual generationOurs · Counterfactual, ImageNet-1k, SD1.5CLIPScore: logit 24.875 -> 23.372; CLIPScore 0.1908 -> 0.1936Mixed
trajectory control vs FreeDoM in text-guided editingOurs · CelebA-HQ, SD1.5ImageReward: -0.1100 -> 0.0976Improved
same method on flow-matching StableDiffusion 3Ours · Human preference, SD3ImageReward: 1.5887 -> 1.8529 vs best inversion baselineImproved
use noiseless reverse path instead of Markovian stochastic initial trajectory (ablation)Ours · Initial trajectory choicestability to hyperparametersImproved
increase compute for baselines (more N_recur and N_iter) (ablation)Inversion-based baselines · Human preferencerewardHurt
trajectory control vs baselines in user studyOurs · User studyreward alignment rating: best baseline 3.28 -> 3.67Improved

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