AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Advantage Weighted Matching: Aligning RL with Pretraining in Diffusion Models

Shuchen Xue, Chongjian GE, Shilong Zhang, Yichen Li, Zhi-Ming Ma

ICML 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning · 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
use clean-target flow matching with advantage weights instead of DDPO per-step likelihoodAWM vs Flow-GRPO · GenEval Overall (SD3.5M)overall score: 0.95 -> 0.985; 16.54x speed-upImproved
AWM instead of Flow-GRPO on OCRSD3.5M · OCR (SD3.5M)accuracy: 0.89 -> 0.92 with 19.13x speed-upImproved
AWM instead of Flow-GRPO on PickScoreSD3.5M · PickScore (SD3.5M)PickScore: 23.01 -> 23.11 with 34.67x speed-upImproved
AWM instead of Flow-GRPO on FLUX OCRFLUX · OCR (FLUX)accuracy: 0.95 -> 0.95 with 8.53x speed-upImproved
AWM instead of Flow-GRPO on FLUX PickScoreFLUX · PickScore (FLUX)PickScore: 23.08 -> 23.08 with 6.82x speed-upNo clear effect
train AWM longer on SD3.5M OCRAWM · OCR (SD3.5M)accuracy: 0.92 -> 0.96Improved
use 6 Monte Carlo timesteps instead of 4 to estimate the AWM objective (ablation)AWM · GenEval (SD3.5M)GenEval scoreImproved
on-policy rollout instead of EMA rollout with importance sampling (ablation)AWM · GenEval (SD3.5M)GenEval scoreNo clear effect
increase KL coefficient beta from 1 to 3 (ablation)AWM · GenEval (SD3.5M)GenEval scoreHurt
SDE sampling instead of ODE sampling during AWM training (ablation)AWM · OCR and ImageRewardrewardNo clear effect
noisy-target denoising score matching instead of clean-target in pretraining (ablation)EDM · CIFAR-10 and ImageNet-64FIDHurt

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