Oriel — AI & technology worth noticing, every day.

NorthCUA Technical Report: Turbocharging Computer-Use Agents through Experience at Scale

Paper · https://arxiv.org/abs/2611.02102

Scaling computer-use agents from individual actions to complex software work requires experience that connects decisions to complete outcomes. We present NorthCUA, an integrated system for turbocharging computer-use agents through experience at scale. Its Experience Flywheel supplies diverse executable tasks to the NorthCUA Harness, which sustains long interactions, records decisions, and checks task outcomes. The Flywheel uses these records to curate verified trajectories. For unsolved tasks, recursive self-improvement evolves reusable skills and preserves verified milestones so later attempts build on earlier progress. Staged Learning combines this experience with grounding data from the Flywheel to develop GUI competence through mid-training and long-horizon supervised fine-tuning. The policy then learns from fresh attempts in the same Harness through joint desktop-web reinforcement learning, closing the loop between experience production and policy improvement. The resulting Nano, Mini, and Pro family achieves state-of-the-art results among open-source models across GUI grounding, desktop execution, and Web interaction under the reported evaluation settings. For more responsive Web workflows, NorthCUA-Ultrafast pairs direct action selection with open-ended generation on pre-RL Nano, offering a measured quality-latency tradeoff. Together, these advances establish NorthCUA as a high-performing model family and a reusable foundation for agents that carry complex software work from intent to completion.

First seen by Oriel on .

AI summary

NorthCUA is a computer-use agent system that uses large-scale task experience and iterative learning to improve performance across GUI grounding, desktop execution, and web interaction.

  • Its Experience Flywheel supplies executable tasks to a Harness that records decisions and verifies outcomes, then curates the resulting trajectories.
  • For unsolved tasks, recursive self-improvement develops reusable skills and retains verified milestones for future attempts.
  • Staged Learning combines Flywheel experience and grounding data through mid-training and long-horizon supervised fine-tuning; joint desktop-web reinforcement learning then uses fresh Harness attempts.
  • The Nano, Mini, and Pro models report state-of-the-art results among open-source models under the stated evaluation settings.
  • NorthCUA-Ultrafast pairs direct action selection with open-ended generation on pre-RL Nano to offer a measured quality-latency tradeoff for web workflows.

#computer-use agents #reinforcement learning #gui agents #web automation

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Points

Score reported by the source.

  • Hugging Face Papers: 32, latest observation ; +2 over observed period since

Comments

Comment count reported by the source.

  • Hugging Face Papers: 1, latest observation ; 0 over observed period since

Aliases

SourceExternal IDURL
Hugging Face Papers2611.02102https://arxiv.org/abs/2611.02102 Read discussion