AI Paper Atlas

Research map / Video / Video generation

GenFlowRL: Shaping Rewards with Generative Object-Centric Flow in Visual Reinforcement Learning

Kelin Yu, Sheng Zhang, Harshit Soora, Furong Huang, Heng Huang, Pratap Tokekar, Ruohan Gao

ICCV 2025 · Video › Video generation · approach: World-model video generation · 5 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 GenFlowRL instead of Im2Flow2Act imitation policyGenFlowRL · Im2Flow2Act Pivoting (demo-conditioned)success rate (%): 60 -> 90Improved
use GenFlowRL instead of Im2Flow2Act imitation policyGenFlowRL · Im2Flow2Act Folding (language-conditioned)success rate (%): 35 -> 80Improved
use GenFlowRL instead of Im2Flow2Act imitation policyGenFlowRL · Im2Flow2Act Pivoting (language-conditioned)success rate (%): 45 -> 85Improved
add large Gaussian noise (x4) to generated flow (ablation)GenFlowRL · Im2Flow2Act Folding (language-conditioned)success rate (%): 80 -> 75No clear effect
add large drift noise (Gauss x2 Drift x2) to generated flow (ablation)GenFlowRL · Im2Flow2Act Pouring (language-conditioned)success rate (%): 95 -> 75Hurt
add large drift noise (Gauss x2 Drift x2) to generated flow (ablation)GenFlowRL · Im2Flow2Act Folding (language-conditioned)success rate (%): 80 -> 65Hurt
encode keypoints with delta-flow instead of MLP (ablation)GenFlowRL · PickNPlace, Pouring, Pivoting (RL training)success rateImproved
add initial 3D centroids to policy input (ablation)GenFlowRL · PickNPlace, Pouring, Pivoting (RL training)success rateImproved
use 128 keypoints instead of 64 or 32 (ablation)GenFlowRL · PickNPlace, Pouring, Pivoting (RL training)success rateNo clear effect
use delta-flow reward instead of VIPER, Diffusion Reward, RND or sparse rewardGenFlowRL · MetaWorld (5 hard tasks)training success rateImproved

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