WorldLine: 10k-hour video-trained action simulator lifts robot policy success up to 21.4 pts
hongkongust · hf · 2026-09-30
WorldLine is an action-driven visual simulator that predicts manipulation outcomes before physical execution.
- Decouples transferable dynamics learning from heterogeneous action grounding: 10,000+ hours of action-free robot video for dynamics, 2,000+ hours of trajectories across 10+ embodiments for grounding
- Image-space action representation gives a shared control interface across embodiments; multi-view, failure-enriched training and few-step distillation improve interaction-sensitive causal rollout
- Improves robot-mask IoU by 0.1626 on failed trajectories; predicts trajectory success at 74% mean accuracy across RoboTwin and AgiBot
- Without RoboTwin training or adaptation, its rollouts improve task success by up to 21.4 percentage points over direct policy execution
More from Embodied
- DIY open-source driving mods: Tesla owner runs Sunnypilot on Model Y with $1,000 Comma Four kit, alarming experts — science · 2026-09-30
- Call for physical AI founders: actuators, dexterous hands, robot foundation models and more — Sethwinterroth · 2026-09-30
- EVO-WAM self-verifies video-action rollouts, real-world long-horizon success 20%→76.7% — hitsz123 · 2026-09-30
- PanoVLN beats VLN SOTA by 11.9% success rate using panoramic views, tested on a quadruped robot — ZhejiangUniversity · 2026-09-30
- Agile Robots ships humanoid robots daily from German factory, partners with DeepMind — CyberRobooo · 2026-09-30
- Real2Gym Turns Videos Into Interactive Robot Gyms, Beating GPT-6 Direct Mode by 33% on Real Robots — Shanghai-AI-Laboratory · 2026-09-30