iSHIFT: Lightweight Slow-Fast GUI Agent with Adaptive Perception
Sarthak Mehrotra , Sairam V C Rebbapragada , Mani Hemanth Reddy Bonthu , Vineeth N Balasubramanian
- 🏛 Institutions
- Indian Institute of Technology , Bombay , Indian Institute of Technology , Hyderabad
- 📅 Date
- December 26, 2025
- 📑 Publisher
- arXiv
- 💻 Env
- General GUI
- 🔑 Keywords
TLDR
iSHIFT is a 2.5B GUI agent that combines latent thinking with perception-control tokens so it can switch between a fast global mode and a slower grounding-heavy mode. The paper positions this as a way to allocate reasoning depth and visual focus adaptively while still matching state-of-the-art results on multiple GUI benchmarks.
Related papers (24)
- CocoaBench: Evaluating Unified Digital Agents in the WildApril 13, 2026 · arXiv
- Are GUI Agents Focused Enough? Automated Distraction via Semantic-level UI Element InjectionApril 9, 2026 · arXiv
- ToolTok: Tool Tokenization for Efficient and Generalizable GUI AgentsJanuary 30, 2026 · arXiv
- Visual Grounding for User InterfacesJune 16, 2024 · NAACL 2024 Industry Track
- How do Visual Attributes Influence Web Agents? A Comprehensive Evaluation of User Interface Design FactorsJanuary 29, 2026 · arXiv
- Iris: Breaking GUI Complexity with Adaptive Focus and Self-RefiningDecember 13, 2024 · arXiv
- Dual-View Visual Contextualization for Web NavigationFebruary 6, 2024 · CVPR 2024 (Poster)
- Naive Visual Memory is Not Enough: A Failure-Mode Study of GUI AgentsJune 12, 2026 · arXiv
- Demo2Tutorial: From Human Experience to Multimodal Software TutorialsJune 2, 2026 · arXiv
- STaR-KV: Spatio-Temporal Adaptive Re-weighting for KV Cache Compression in GUI Vision-Language ModelsJune 1, 2026 · arXiv
- GUI-C²: Coarse-to-Fine GUI Grounding via Difficulty-Aware Reinforcement LearningMay 29, 2026 · arXiv
- MementoGUI: Learning Agentic Multimodal Memory Control for Long-Horizon GUI AgentsMay 18, 2026 · arXiv
- Video2GUI: Synthesizing Large-Scale Interaction Trajectories for Generalized GUI Agent PretrainingMay 14, 2026 · arXiv
- Executable Agentic Memory for GUI AgentMay 12, 2026 · arXiv
- LiteGUI: Distilling Compact GUI Agents with Reinforcement LearningMay 8, 2026 · arXiv
- Step-level Optimization for Efficient Computer-use AgentsApril 29, 2026 · arXiv
- Training Computer Use Agents to Assess the Usability of Graphical User InterfacesApril 28, 2026 · arXiv
- AutoGUI-v2: A Comprehensive Multi-Modal GUI Functionality Understanding BenchmarkApril 27, 2026 · arXiv
- Human-Guided Harm Recovery for Computer Use AgentsApril 20, 2026 · arXiv
- UI-Zoomer: Uncertainty-Driven Adaptive Zoom-In for GUI GroundingApril 15, 2026 · arXiv
- GUI-Perturbed: Domain Randomization Reveals Systematic Brittleness in GUI Grounding ModelsApril 15, 2026 · arXiv
- See, Point, Refine: Multi-Turn Approach to GUI Grounding with Visual FeedbackApril 14, 2026 · arXiv
- What's Missing in Screen-to-Action? Towards a UI-in-the-Loop Paradigm for Multimodal GUI ReasoningApril 8, 2026 · Findings of ACL 2026
- MAESTRO: Adapting GUIs and Guiding Navigation with User Preferences in Conversational Agents with GUIsApril 7, 2026 · arXiv