← back to paper
arxiv: 2607.09776 · 2 revisions
HELP: Human-Efficient Large-Scale Robot Post-Training with Rollout Segmentation