Reproduce “Work Agents: Systems, Capabilities, and Optimization” 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
Zishan Xu, Enjun Du, Mingqi Gao, Cheng Yang, Junhao Shen, Zirong Chen, Zhiyuan Yao, Yifu Guo and 8 more
The authors propose a framework for understanding work agents as systems accountable for verified outcomes rather than models producing answers or tool calls. It combines a model, harness, and workspace, and distinguishes persistence—when responsibility can close—from openness—what must be discovered. The survey synthesizes five work forms and capability families but does not establish improved performance or reliability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.