ScopeMem: Latent Memory That Learns Where to Revise Frozen GUI Agents
Zhicheng Zheng , Cheng Zhang , Shengping Yu , Tianxiang Zhao
- 🏛 Institutions
- HKUST , Imperial College London
- 📅 Date
- October 19, 2026
- 📑 Publisher
- arXiv
- 💻 Env
- Mobile General GUI
- 🔑 Keywords
TLDR
Introduces ScopeMem, a latent memory mechanism for GUI agents that identifies which layers of a frozen vision-language model benefit from history prefix vectors, avoiding uniform token injection and improving task completion on AndroidWorld.
Related papers (24)
- From Interaction to Capability: Task Advancement as State and Supervision for GUI AgentsOctober 27, 2026 · arXiv
- AnchorGUI: Asymmetric Memory for Dual-Scale Learning in GUI NavigationSeptember 14, 2026 · ECCV 2026
- SE-GA: Memory-Augmented Self-Evolution for GUI AgentsMay 16, 2026 · ICML 2026
- Beyond Next-Screen Prediction: The Decision Value of GUI World ModelsOctober 17, 2026 · arXiv
- Where to Learn and How Far to Roll Out: Efficient Adaptive Training for GUI AgentsOctober 29, 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
- Focused Omni-Modal GUI AgentsOctober 19, 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
- Latent World Action Model for GUI AgentsOctober 9, 2026 · arXiv
- DeltaReplay: Task-Relative Memory Reuse for Mobile GUI AgentsOctober 8, 2026 · arXiv
- Not All Experience Belongs in the Weights: Component Routing for Self-Improving GUI AgentsOctober 1, 2026 · arXiv
- PrecogUI: Proactive GUI Agents via Pre-cognitive Simulation and Experience RetrievalSeptember 29, 2026 · arXiv
- PhoneCLI: From App Interfaces to Callable Commands for Mobile AgentsSeptember 28, 2026 · arXiv
- Relevance Does Not Imply Applicability: Experience Activation for Personal GUI AgentsSeptember 27, 2026 · arXiv
- The GUI Is Not the State: Diagnosing State Aliasing in GUI World ModelsSeptember 26, 2026 · arXiv
- PastForward: Faster On-Device GUI Agents via Computational Experience ReuseSeptember 26, 2026 · arXiv
- BlueLM-GUI Technical Report: A Real-Device-Centric Flywheel for Self-Improving Mobile GUI AgentsSeptember 11, 2026 · arXiv
- Efficient GUI Agents: A Systems Survey of Observation, Memory, Action, and Runtime OptimizationSeptember 2, 2026 · arXiv
- Are Android GUI Agents Robust Against Runtime Anomalies? AnTrap: Evaluating Agents in Dynamic Adversarial EnvironmentsAugust 25, 2026 · arXiv
- CausalCache: Conditional High-Fidelity Restoration for Long-Horizon GUI AgentsAugust 23, 2026 · arXiv
- Executable Agentic Memory for GUI AgentMay 12, 2026 · arXiv
- UI-Voyager: A Self-Evolving GUI Agent Learning via Failed ExperienceMarch 25, 2026 · arXiv