EmbodiedSmith: Scaling Embodied Data through Recursive Self-Improvement Flywheel in Simulation

Key points

• Scaling robotic foundation models requires diverse training data and reliable evaluation environments.1
• Simulation offers a scalable solution, yet existing generation pipelines remain constrained by predefined assets and skills, a disconnect between scene generation and task generation, and limited support for complex embodiments and physics.1
• EmbodiedSmith unifies asset, scene, and task generation in a pipeline that supports autonomous creation and language-driven customization.1
• Its core is an agentic refinement loop: scene generation anticipates downstream task requirements, while task generation guides targeted scene edits, allowing scenes and tasks to iteratively improve one another.1

Sources (1)

• [1]EmbodiedSmith: Scaling Embodied Data through Recursive Self-Improvement Flywheel in Simulation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 08:34 AM

  We introduce EmbodiedSmith, a framework for scalable embodied data generation through recursive self-improvement (RSI).

  Scaling robotic foundation models requires diverse training data and reliable evaluation environments.

Extractive summary: sentences quoted from the sources.