EmbodiedRSI: Active Continual Robot Learning Through Hypothesis-Guided Co-Evolution

Key points

• Robot foundation models provide strong visuomotor control, yet their performance can degrade when object positions or task instructions change.1
• Agentic harnesses can adapt around the model, but current self-evolving harnesses use robot trials inefficiently when deciding which code and skill changes to pursue.1
• EmbodiedRSI realizes this through a Fast-Slow Dual-System Architecture, in which competing code and skill hypotheses are maintained in a Hypothesis Graph.1
• The Slow System builds Hierarchical Memory, and Reward-Grounded Memory Learning selects effective memory according to their value for later Fast-System improvement.1

Sources (1)

• [1]EmbodiedRSI: Active Continual Robot Learning Through Hypothesis-Guided Co-Evolution
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 05:48 PM

  We introduce EmbodiedRSI, a self-evolving agentic harness that autonomously decides where to explore next and turns the resulting physical interaction into improved code and skills.

  Robot foundation models provide strong visuomotor control, yet their performance can degrade when object positions or task instructions change.

Extractive summary: sentences quoted from the sources.