Seed2Scale: A Self-Evolving Data Engine with Parallel Worlds Expansion for Scalable Robot Learning
Seed2Scale: A Self-Evolving Data Engine with Parallel Worlds Expansion for Scalable Robot Learning
This paper presents Seed2Scale, a self-evolving data engine designed to overcome challenges in robot learning, such as limited exploration and domain shifts. Starting with minimal seed demonstrations, Seed2Scale uses a combination of lightweight models for data collection and verification, followed by parallel worlds expansion to generate diverse data. Experimental results show strong scaling potential, with a 75.38% success rate in zero-shot real-world evaluations.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Reported success rate
75.38%
Reported zero-shot result; the excerpt does not specify an evaluation protocol.
View original evidence
achieves a remarkable 75.38% success rate in zero-shot real-world evaluations
Open source S5
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
at Seed2Scale exhibits significant scaling potential: as iterations progress, the success rate of the target model shows a consistent upward trend, significantly outperforming the seed baseline. Notably, Seed2Scale achieves a remarkable 75.38% success rate in zero-shot real-world evaluations, where baseline methods fail completely (0%). Project page: https://terminators2025.github.io/Seed2Scale.github.io
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org