Synthoid.ai

“RDGen leverages trained RL policies as a structured trajectory generator... Successful rollouts are then harvested as clean, high-quality demonstrations for downstream VLA training.”

“arXiv:2605.30957v1 [cs.RO] 29 May 2026, Synthoid.ai”

“We train a π₀ policy using the collected trajectories and report the success rate over 10 trials.”

“We train the policy in Isaac Sim, where the robot embodiment, workspace, and task objects are modeled.”