Reproduce “Fewer Tokens, Better Action: GPT-6 Astra Robot Agents with 14% Higher Success Rate but 65% Fewer Tokens” 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
Ruiyang Si, Jianxin Bi, Shunyu Yang, Rui Ni, Wenbo Huang, Qiang Wang, Shulong Jiang, Yintianrun Zhang and 7 more
PyRUA-Lean lets a robot-planning model write Python cells to compose robot operations, check results, retry locally, and return requested outputs and images. In a matched 700-instance simulation using the same GPT-6 Astra planner and primitives, success rose from 63.1% to 71.7%; jointly solved episodes used 49% fewer LLM calls and 65% fewer input tokens. This tests whole interfaces, not individual features.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.