AI Paper Atlas

Research map / Video / Video segmentation and tracking

Spatiotemporal Predictive Pre-training for Robotic Motor Control

Jiange Yang, Bei Liu, Jianlong Fu, Bocheng Pan, Gangshan Wu, Limin Wang

IJCV 2026 · Video › Video segmentation and tracking · approach: Physics-inspired motion prediction · 28 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
add temporal prediction with future frame (STP vs MAE)MAE pre-training · Single-task simulationweighted average: 59.6 -> 63.7Improved
add ImageNet-MAE hybrid initializationSTP · Single-task simulationweighted average: 63.7 -> 64.2Improved
add temporal prediction (STP vs MAE)MAE pre-training · LIBERO-LONGsuccess rate: 37.3 -> 41.3Improved
add temporal prediction (STP vs MAE)MAE pre-training · RLbenchsuccess rate: 45.2 -> 51.3Improved
use STP post-pre-training instead of MAE post-pre-trainingpost-pre-training · Single-task simulationweighted average: 72.5 -> 76.4Improved
end-to-end fine-tuning instead of frozen encoderSTP · Single-task simulationweighted average: 63.7 -> 62.9Hurt
scale ViT-B to ViT-L with post-pre-trainingSTP · Single-task simulationweighted average: 76.4 -> 78.4Improved
STP vs VC-1 in real world standard environmentSTP · Real-worldsuccess rate: 42.0 -> 49.3Improved
STP vs VC-1 in unseen environment with distractorsSTP · Real-worldsuccess rate: 36.7 -> 46.3Improved
add spatial prediction on the 75% masked current frame (ablation)STP · Single-task simulationweighted average: 57.4 -> 63.7Improved
use future frame 95% masked condition instead of language only (ablation)STP · Single-task simulationweighted average: 55.4 -> 63.7Improved
add language narration to 95% masked future frame condition (ablation)STP · Single-task simulationweighted average: 63.7 -> 63.1No clear effect

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