Reproduce “ScienceCUA: Resource-to-Skill Construction and Execution-Guided Skill Evolution for Scientific Computer Use Agents” 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
Daiqiang Li, Jingyi Peng, Rui Hu, Jia Li, Xuyang Liu, Guanren Qiao, Yujia Xiang, Qiuzhuang Sun and 2 more
ScienceCUA builds a library from software manuals, separating reusable workflow Skills from shared lookup References. During execution, it uses failures to revise guidance or add source-supported knowledge, validating candidate updates in fresh sessions before freezing the library. On a six-application benchmark, Grok-4.6’s unweighted average rose from 37.47% to 54.59%; held-out evidence covers 168 tasks, not all agents or software.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.