Reproduce “VABAR: A Variational Automation Benchmark for Agentic Robotics” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shuangyu Xie, Kaiyuan Chen, Uksang Yoo, Xinyue Zhu, Meadow Shen, Jaimyn Drake, Eugen Solowjow, Lionel Gulich and 11 more
VABAR evaluates agentic robot programs across six simulated tasks with structured, physically feasible variations. It uses reusable perception, planning, and control skills, and records success and diagnostics. On Gear Meshing Level 1, autonomously generated GaP succeeds on 70.0% of instances, versus 0.0% for Claude Code and 100.0% for a human-authored graph; these results are simulated, not evidence of real-world operation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.