AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Lookahead Sample Reward Guidance for Test-Time Scaling of Diffusion Models

Yeongmin Kim, Donghyeok Shin, Byeonghu Na, Minsang Park, Richard Lee Kim, Il-chul Moon

ICML 2026 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 0 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
replace gradient guidance (DATE) with LiDAR (DPM-8, n=50)SDXL · GenEval promptsGenEval: 0.570 -> 0.585Improved
use a distilled one-step DMD model as lookahead sampler (ablation)LiDAR on SDXL · GenEval promptsGenEval: 0.585 -> 0.598Improved
increase lookahead samples n from 3 to 50 (DPM-5) (ablation)LiDAR on SD v1.5 · GenEval promptsGenEval: 0.449 -> 0.478Improved
add LiDAR on top of DPO fine-tuningDPO-tuned SD v1.5 · GenEval promptsGenEval: 0.446 -> 0.489Improved
use Universal Guidance (gradient) with ImageRewardSD v1.5 · GenEval promptsGenEval: 0.426 -> 0.355Hurt
add negative guidance from low-reward lookahead samples (Safe-D, SR) (ablation)SD v1.5 · ImageRewardImageReward: 0.001 -> 0.014No clear effect
increase vanilla sampling steps from 100 to 250SD v1.5 vanilla · GenEval promptsGenEval: 0.426 -> 0.430No clear effect
mix 10% CLIP into the IR reward annotation (ablation)LiDAR · SD v1.5 GenEval promptsGenEval: 0.478 -> 0.473Mixed
use CLIP only as reward instead of IR (ablation)LiDAR · SD v1.5 GenEval promptsGenEval: 0.478 -> 0.454Hurt
increase lookahead samples n on FLUX (ablation)LiDAR on FLUX · GenEvalGenEval: 0.645 -> 0.668Improved
increase lookahead samples n on discrete diffusion (ablation)LiDAR on UDLM (QM9) · ring count rewardnovel ring count: 2.192 -> 4.128Improved
add LiDAR to SMC, BoN and SMC+BoNSMC / BoN · SD v1.5 ImageReward, N=64ImageReward: 1.221 -> 1.338 (SMC)Improved

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