# RoboSTD researchers report two-arm learning gains from single-arm robot data

_Published Tuesday, October 27, 2026 at 9:28 PM EDT · Science, Robotics · Latest · Tier 2 — Notable_

Researchers report that RoboSTD, a method for training two-arm robots using single-arm demonstrations, increased the average number of cups collected to 2.11x the single-arm baseline without additional two-arm demonstrations. The arXiv preprint describes evaluations on the AgileX ALOHA platform in simulation and real-world settings.

The method mirrors observations, robot-state information and actions across a plane dividing left from right to create demonstrations for the opposite arm. A language model then generates spatial and timing constraints to coordinate the arms. The reported numerical gain concerns a cup collection task.

## Sources

- [arXiv Query: search_query=cat:cs.RO&id_list=&start=0&max_results=30](https://arxiv.org/abs/2610.36685v1)

---
Canonical: https://techandbusiness.org/newswire/IBSSzfC_inQwj78MINe8ZQ
Published: 2026-10-28T01:28:29.960Z
Story chronology: 2026-10-27T05:56:15.000Z
Retrieved: 2026-10-28T03:39:19.223Z
Publisher: Tech & Business (techandbusiness.org)