AI Paper Atlas

Research map / Vision-language and multimodal / Robot action and navigation

IRASim: A Fine-Grained World Model for Robot Manipulation

Fangqi Zhu, Hongtao Wu, Song Guo, Yuxiao Liu, Chilam Cheang, Tao Kong

ICCV 2025 · Vision-language and multimodal › Robot action and navigation · approach: Memory and subgoal-based manipulation · 119 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 frame-level instead of video-level action conditioning (ablation)IRASim · RT-1 shortPSNR: 25.446 -> 26.048Improved
use frame-level instead of video-level action conditioning (ablation)IRASim · Bridge shortPSNR: 24.733 -> 25.275Improved
use frame-level instead of video-level action conditioning (ablation)IRASim · Language-Table shortPSNR: 23.893 -> 28.818Improved
use frame-level instead of video-level action conditioning (ablation)IRASim · Language-Table long trajectoriesLatent L2: 0.2112 -> 0.1730Improved
use frame-level instead of video-level action conditioning (ablation)IRASim · RT-1 shortFID: 4.34 -> 5.60Hurt
use 2 context frames instead of 1 (ablation)IRASim · BridgePSNR: 25 -> 25No clear effect
train world model with post-trained rollouts (P=1000 vs P=0) (ablation)IRASim planning · Push-T K=50IoU: 0.418 -> 0.961Improved
sample more trajectories (K=50 vs K=1) with P=1000 (ablation)IRASim planning · Push-TIoU: 0.637 -> 0.961Improved
sample more trajectories (K=50 vs K=1) without post-trained rollouts (P=0) (ablation)IRASim planning · Push-TIoU: 0.637 -> 0.418Hurt
increase post-trained rollouts from 500 to 1000 at K=5 (ablation)IRASim planning · Push-T K=5IoU: 0.907 -> 0.886Hurt
use MSE instead of ResNet50 cosine similarity as value function (ablation)IRASim planning · Real-robot Close Drawersuccess rate: 0.60 -> 0.87Improved
use IRASim planning (MSE) instead of random trajectory selectionIRASim planning · Real-robot Close Drawersuccess rate: 0.20 -> 0.87Improved

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