EmbodiedSmith: Recursive Self-Improvement Flywheel Scales Embodied Training Data in Simulation
Yikai Qin · hf · 2026-10-07
EmbodiedSmith is a framework for scalable embodied data generation via recursive self-improvement (RSI) in simulation, addressing bottlenecks in robotic foundation model training.
- Existing pipelines suffer from predefined assets/skills, disconnection between scene and task generation, and limited support for complex embodiments and physics;
- It unifies asset, scene, and task generation with autonomous creation and language-driven customization;
- An agentic refinement loop lets scene generation anticipate task requirements while task generation drives targeted scene edits, iteratively improving both—including long-horizon tasks;
- Supports mobile manipulators, humanoids, dexterous hands, deformable objects, and fluids.
Experiments validate data quality, diversity, and efficiency; downstream policies show improved generalization with more diverse data.
More from Embodied
- Minerva Humanoids debuts Roger robot for dangerous jobs, raises $10M pre-seed — Scobleizer · 2026-10-07
- 7 minutes per motor, ~$5 labor cost: why robotics automation is the only path for Western manufacturing — avlok · 2026-10-07
- Reward-DAgger: generalist reward models enable task-agnostic runtime monitoring for robots — ebiyik_ · 2026-10-07
- NVIDIA's VeriFine Co-Evolves Policy and Judge to Scale Self-Improvement in Embodied Reasoning — nvidia · 2026-10-07
- ETH student documents building and calibrating a UMI gripper from scratch — SongShuran · 2026-10-07
- Kangwook Lee on how games can help build AI for the physical world — Kangwook_Lee · 2026-10-07