LeapBot-JAC: Joint Action Chains for World Action Models

Key points

• Action-chunking policies predict future controls but execute only a prefix before replanning.1
• How should a fixed action-token budget be organized to support execution?1
• Both chains are supervised by demonstrated controls and jointly denoised at aligned physical-time positions.1
• JAC adds neither action tokens nor parameters and retains the host's world-modeling process; both chains participate in inference, while only the execution chain drives the robot.1

Sources (1)

• [1]LeapBot-JAC: Joint Action Chains for World Action Models
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 02:21 AM

  We propose LeapBot-JAC, which allocates sparse future-control references and dense execution targets within one shared action expert.

  Action-chunking policies predict future controls but execute only a prefix before replanning.

Extractive summary: sentences quoted from the sources.