Reproduce “NorthCUA Technical Report: Turbocharging Computer-Use Agents through Experience at Scale” 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
Chaoyun Zhang, Bingxuan Li, Pinxue Guo, Yangfan Gao, Zikun Cai, Tianci Xie, Hui Li, Yuqi Huang and 22 more
NorthCUA trains computer-use agents through stateful desktop and web tasks whose outcomes are checked, then uses selected experience and fresh attempts to update the policy. Its harness preserves trajectories and workspaces, while staged learning moves from supervised fine-tuning to joint reinforcement learning. NorthCUA-Pro reaches 73.9% on WebArena-Verified, a benchmark result rather than a general measure of computer-use ability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.