AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Distillation Models are Good Samplers for Diffusion Reinforcement Learning

Zunxu Liu, Aiqiu Wu, Zhaofan Qiu, Yingwei Pan, Ting Yao, Tao Mei

ICML 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
use late distillation sampling (S3) instead of static fully distilled sampling (S2) (ablation)RL with distilled sampler · Image OCROCR accuracy: 0.79 -> 0.95Improved
use late distillation sampling (S3) instead of early distillation sampling (S4) (ablation)RL with distilled sampler · Image OCROCR accuracy: 0.81 -> 0.95Improved
use late distillation sampling (S3) instead of static fully distilled sampling (S2) (ablation)RL with distilled sampler · Video OCROCR accuracy: 0.26 -> 0.42Improved
replace vanilla 40-step sampling (S1) with late distillation sampling (S3) (ablation)RL with distilled sampler · Image OCRtraining time (GPU hours): 160 -> 16Improved
add reward-aware distillation (ablation)DMSampler · Image OCROCR accuracy: 0.96 -> 0.97Improved
add reward-aware distillation (ablation)DMSampler · Video OCROCR accuracy: 0.48 -> 0.50Improved
add iterative co-evolution phases (RL phase 1 to 3) (ablation)DMSampler · Image OCRpolicy OCR accuracy: 0.95 -> 0.97Improved
add iterative co-evolution phases (RL phase 1 to 3) (ablation)DMSampler · Video OCRpolicy OCR accuracy: 0.42 -> 0.50Improved
use distilled sampler instead of direct step reduction with CFG (DRS3, 10 steps)DMSampler · Video OCROCR accuracy: 0.23 -> 0.50Improved
use DMSampler instead of DiffusionNFT with 50-step CFG samplingDiffusionNFT · VBenchtotal score: 84.74 -> 85.00Improved
add DMSampler to FlowGRPOFlowGRPO · GenEvalGenEval score: 0.95 -> 0.96Improved
add DMSampler to DGPODGPO · GenEvalGenEval score: 0.97 -> 0.97No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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