Mobile-Agent-v2: Mobile Device Operation Assistant with Effective Navigation via Multi-Agent Collaboration
Junyang Wang , Haiyang Xu , Haitao Jia , Xi Zhang , Ming Yan , Weizhou Shen , Ji Zhang , Fei Huang , Jitao Sang
- 🏛 Institutions
- Beijing Jiaotong University , Alibaba Group
- 📅 Date
- June 3, 2024
- 📑 Publisher
- NeurIPS 2024
- 💻 Env
- Mobile
- 🔑 Keywords
TLDR
Mobile-Agent-v2 is a mobile operation assistant that decomposes control into planning, decision, and reflection agents to handle long history navigation more effectively. It also maintains a memory unit for focus content from prior screens and reports over 30% task-completion improvement over the earlier single-agent Mobile-Agent setup.
Related papers (24)
- MobileExperts: A Dynamic Tool-Enabled Agent Team in Mobile DevicesJuly 4, 2024 · arXiv
- Seeing is Believing: Vision-driven Non-crash Functional Bug Detection for Mobile AppsJuly 3, 2024 · arXiv
- World-Model-Augmented Web Agents with Action CorrectionFebruary 17, 2026 · arXiv
- Online Dense Reinforcement Learning with Planning-Grounding Agent for UI Navigation on Live IOS AppsOctober 29, 2026 · arXiv
- GUI-JEPA: Candidate-Conditioned JEPA Latents as a Fast Action Policy for Mobile GUI AgentsOctober 27, 2026 · arXiv
- iPrefBench: A User-Centric Benchmark for Personalized Mobile Use AgentsOctober 26, 2026 · arXiv
- Chart the Invisible: Interactive Perception for Hidden-Target Exploration in GUI AgentsOctober 26, 2026 · arXiv
- AndroidFlux: Evaluating Failure Recovery and Recovery Action Reward Modeling in Mobile-Use Multimodal AgentsOctober 25, 2026 · arXiv
- Subgoal-aware Memory Graph for GUI AgentOctober 22, 2026 · arXiv
- Reuse, Rewrite, or Reject: Evidence-Constrained Trace Reuse for Repeated Mobile GUI ControlOctober 22, 2026 · arXiv
- Diagnosing Personalization Failures in Mobile GUI AgentsOctober 22, 2026 · arXiv
- Tandem: Serving Hybrid Linear-Attention GUI Agents on Mobile NPUsOctober 19, 2026 · arXiv
- ScopeMem: Latent Memory That Learns Where to Revise Frozen GUI AgentsOctober 19, 2026 · arXiv
- Beyond Next-Screen Prediction: The Decision Value of GUI World ModelsOctober 17, 2026 · arXiv
- MobileJailBench: Red Queen Jailbreak Benchmarking for Mobile AgentsOctober 16, 2026 · arXiv
- MobileInject: Evaluating Reasoning-to-Action Security of MLLM-Based Mobile Agents under Cross-Modal Consistency InjectionOctober 16, 2026 · arXiv
- Not All Graphs Are Maps: A Task-Agnostic Exploration Framework for Mobile GUI NavigationOctober 15, 2026 · arXiv
- MAI-Gym: An Open All-in-One Platform for GUI Agent Evaluation and RL Training across Mobile and Computer EnvironmentsOctober 14, 2026 · arXiv
- How GUI Agents Fail: GraceBench for Execution Stability DiagnosisOctober 14, 2026 · arXiv
- Learning How to Remember: Memory Modality Selection for GUI AgentsOctober 13, 2026 · arXiv
- Adaptive-Group Process Rewards for GUI AgentsOctober 13, 2026 · arXiv
- Seg2Click: Learning Pixel-Level Action Masks for GUI Visual GroundingOctober 12, 2026 · arXiv
- PrivacyGuard: Source-Linked Sanitization and Locally Resolvable PII References for Cloud GUI AgentsOctober 12, 2026 · arXiv
- OpenMobile-2: Building Versatile Mobile Agents with Scalable Environments and App-Native ToolsOctober 9, 2026 · arXiv