AI
HybridCUA: Learning to Orchestrate GUI and CLI for Computer-Use Agents
Publication: Hugging Face Daily Papers
Author: Not available in collected metadata
Source published: 28/09/2026, 20:00:00 UTC
Briefing created: 30/09/2026, 04:37:45 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 30/09/2026, 04:18:25 UTC.
Briefings
This article presents HybridCUA, a computer use agent framework that combines GUI and CLI capabilities to improve task performance. The authors developed a data construction pipeline generating hybrid trajectories to train a 9B parameter model, which achieved 53.6% accuracy on OSWorld and 4.0% improvement on WindowsAgentArena compared to base models. The study addresses the challenge of selecting appropriate CLI usage during GUI tasks through supervised fine-tuning and CLI-aware reinforcement learning.
Key points
- The HybridCUA framework combines GUI and CLI interactions to improve computer use agent performance, achieving 53.6% accuracy on OSWorld.
- The authors constructed a data pipeline producing 5K hybrid trajectories and 3K verified RLVR tasks to train the model.
Context and limitations
The study relies on constructed trajectories rather than real-world execution data, and performance metrics are limited to OSWorld and WindowsAgentArena benchmarks.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 30/09/2026, 04:37:45 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeie7ol7gmhvbdcrvhutawqaolu3qgdew3vgvck73xzivde7l6uilha