RoboPlayground: Democratizing Robotic Evaluation through Structured Physical Domains
Published in arXiv preprint, 2026
Recommended citation: Wang, Y. R., Ung, C., Gubarev, E., Tan, C., Srinivasa, S., & Fox, D. (2026). RoboPlayground: Democratizing Robotic Evaluation through Structured Physical Domains. arXiv preprint. https://arxiv.org/abs/2604.05226
A language-driven evaluation framework that lets users author executable robotic manipulation tasks from natural language, compiled into reproducible task specifications. Evaluation on language-defined task families reveals generalization failures not evident in fixed benchmarks, and task diversity scales through crowd contributions.
Authors: Yi Ru Wang, Carter Ung, Evan Gubarev, Christopher Tan, Siddhartha Srinivasa, Dieter Fox