Language Models can Solve Computer Tasks

Studies whether language models can solve computer-control tasks through prompting alone, using Recursive Critique and Improve to iteratively refine action plans. On MiniWoB++, the approach outperforms prior LLM prompting baselines and competitive supervised or RL systems on several tasks.

• A Data-Driven Approach for Learning to Control Computers
  February 16, 2022 · ICML 2022
• JarvisGUI: Towards Cross-Device GUI Agents with Dynamic Task Composition
  September 9, 2026 · arXiv
• FinCUABuild: Can Agents Build Reliable Benchmarks for Dynamic Financial Computer Use?
  September 7, 2026 · arXiv
• AgentHijack: Visual Patch Attacks on Multimodal Computer-Use Agents
  September 6, 2026 · arXiv
• From Interaction Traces to Persistent Skills: Online Evolution for Computer-Use Agents
  September 4, 2026 · arXiv
• CUA-Universe: A Scalable and Dynamic Environment for Hybrid GUI+CLI Agents
  September 4, 2026 · arXiv
• OmegaUse-SOP: SOP Engineering for Professional Computer Use from Human Demonstrations
  September 2, 2026 · arXiv
• Are We There Yet? Assessing Computer-Use Agents for Blind Users' Accessible Interaction with Desktop Applications
  September 1, 2026 · arXiv
• SIR: Self-improving Red-teaming for Compute Use Agents
  August 31, 2026 · arXiv
• CURA: Certified Runtime Alarms for Computer-Use Agents
  August 28, 2026 · arXiv
• ASIL: Replacing Screenshot-and-Click with Structured State and Semantic Actions
  August 27, 2026 · arXiv
• LocalLSTC: A Long Short-Term Control Architecture for Locally Deployed GUI Agents
  August 26, 2026 · arXiv
• Reflection with Action-Induced Visual Differences for Desktop GUI Agents
  August 25, 2026 · arXiv
• ADeptS-Bench: Measuring the Trustworthiness of Computer Use Agents Across Devices
  August 25, 2026 · arXiv
• CONTRAMEM: Learning Self-Evolving Procedural Memory from Contrasting Multi-Model Trajectories
  August 23, 2026 · arXiv
• Spine-Branch Coordination for Multi-agent Computer Use
  August 22, 2026 · arXiv
• Inducing Task Models from Computer-Use Traces
  August 20, 2026 · arXiv
• Screenshots or Tools? Eliciting Tool Use and Managing Multimodal Context in Hybrid GUI-MCP Computer-Use Agents
  August 4, 2026 · arXiv
• Qwen-CUA: Native Computer Use for (almost) Everything
  August 3, 2026 · arXiv
• MAGA: Multi-Platform Self-Fusion of GUI Agents via Structured Action Distillation
  July 31, 2026 · arXiv
• CUADebug: Diagnosing and Repairing Computer-Use Agent Failures
  July 31, 2026 · arXiv
• Qwen-UI-Agent Technical Report: Toward Next-Generation Real-World Centric Foundation GUI Agents
  July 30, 2026 · arXiv
• OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models
  July 30, 2026 · arXiv
• OmegaUse-OfficeVal: Benchmarking LLM Agents on Long-Horizon Office-Suite Tasks with Economic Grounding
  July 29, 2026 · arXiv