AI Paper Atlas

Research map / Video / Video generation

VELR: Efficient Video Reward Feedback via Ensemble Latent Reward Models

Liyu Zhang, Kehan Li, Tao Zhou, Zeyi Huang, Chao Li, Jiming Chen

ICML 2026 · Video › Video generation · approach: Compositional text-to-video diffusion

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 full ReFL with latent reward model (VELR)Wan-2.1 with UnifiedReward · memory during backproptotal memory (GB): 178.59 -> 22.18Improved
use ensemble of 10 LRMs instead of single LRM (ablation)latent reward model · LRM test setMSE: 0.851 -> 0.008Improved
use VELR LRM instead of Dollar LRM (ablation)latent reward model · LRM test setPCC: 0.34 -> 0.91Improved
fine-tune with VELR unified-rewardCogVideoX-1.5 5B · VBench2.0 Averagescore: 52.25 -> 57.16Improved
fine-tune with VELR Causal-VQACogVideoX-1.5 5B · VBench2.0 Aesthetic Qualityscore: 62.53 -> 61.94Hurt
start gradients at step 30 (late) instead of mid (ablation)VELR · VBench Image Qualityscore: 69.62 -> 66.39Hurt
shorten truncated duration to 4 (ablation)VELR · VBench Image Qualityscore: 69.62 -> 66.90Hurt
add online alignment of LRM (ablation)VELR · VBench Image Qualityscore: 68.36 -> 69.62Improved
add online alignment of LRM (ablation)VELR · VBench Aesthetic Qualityscore: 66.34 -> 66.21No clear effect
add uncertainty weighting (ablation)VELR · VBenchscore and GPU hours: GPU hours 15.0 -> 12.6Improved
add KL divergence term (ablation)VELR · qualitative generation diversitydiversityImproved
add truncated mid-step setting vs VADER/ImageReward paradigms (ablation)LRM-based ReFL on Wan-2.1 · qualitativevideo qualityImproved

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