Octopus v3: Technical Report for On-device Sub-billion Multimodal AI Agent

Octopus v3 is a sub-billion multimodal AI agent model designed for efficient on-device deployment, with the paper centered on its functional-token mechanism and edge-device constraints rather than GUI-native interaction. It is relevant to GUI research as a lightweight multimodal agent backbone, but it is broader than a direct GUI-agent paper.

• OS-Helix: Unifying GUI Interaction and Code Execution in Native Computer-Using Agents
  October 15, 2026 · arXiv
• Why Generate a Click? Plug-in Decision Heads for Web Agents
  October 14, 2026 · arXiv
• DeskForge: Dense Supervision from Desktop Environments for Computer-Use Agents
  October 1, 2026 · arXiv
• AutoGUIWorld: Image Generators as Visual World Models for GUI Agent
  October 1, 2026 · arXiv
• HybridCUA: Learning to Orchestrate GUI and CLI for Computer-Use Agents
  September 29, 2026 · arXiv
• CaptchaArena: A Large-Scale, Fine-Grained Dataset for Training Computer-Use Agents on Interactive CAPTCHAs
  September 25, 2026 · arXiv
• Learn How to Act from Your Own Interactions: On-Policy Self-Distillation for GUI Agents
  September 23, 2026 · arXiv
• Towards Omni-dimensional GUI Agent Navigation with Masked Trajectory Prediction
  September 22, 2026 · EMNLP 2026
• MATE: Policy-Aware Security Auditing for Mobile Agents via Synthesis-Driven Trajectory Learning
  September 19, 2026 · USENIX Security 2026
• MintAct: A Unified Visual Agent for Digital Environments
  September 18, 2026 · arXiv
• BlueLM-GUI Technical Report: A Real-Device-Centric Flywheel for Self-Improving Mobile GUI Agents
  September 11, 2026 · arXiv
• LLaDA-UI: Bringing Block-wise Diffusion to Vision-Language GUI Agents
  September 9, 2026 · arXiv
• UI-Venus-2 Technical Report
  August 27, 2026 · arXiv
• Qwen-CUA: Native Computer Use for (almost) Everything
  August 3, 2026 · arXiv