How Far Does Online RL Generalize in Mobile GUI Agents? A Controlled Study with AndroidWorld-Generalization
cs.CV, cs.CL, cs.HC, cs.LG
Submitted: 2026-03-08
Updated: 2026-11-18
Code: https://github.com/zihuanjiang/AndroidWorld-Generalization
Terminology
Sources
- Agent S2: A Compositional Generalist-Specialist Framework for Computer Use Agents
- CausalWorld: A Robotic Manipulation Benchmark for Causal Structure and Transfer Learning
- Qwen2.5-VL Technical Report
- CARL: A Benchmark for Contextual and Adaptive Reinforcement Learning
- A3: Android Agent Arena for Mobile GUI Agents with Essential-State Procedural Evaluation
- Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities
- Group-in-Group Policy Optimization for LLM Agent Training
- Go-Browse: Training Web Agents with Structured Exploration
- Navigating the Digital World as Humans Do: Universal Visual Grounding for GUI Agents
- A Survey on LLM-as-a-Judge
- Mobile-R1: Towards Interactive Capability for VLM-Based Mobile Agent via Systematic Training
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- Contextual Markov Decision Processes
- ScreenQA: Large-Scale Question-Answer Pairs over Mobile App Screenshots
- GPT-4o System Card
- OpenAI o1 System Card
- A Survey of Frontiers in LLM Reasoning: Inference Scaling, Learning to Reason, and Agentic Systems
- Adam: A Method for Stochastic Optimization
- VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks
- MobileUse: A GUI Agent with Hierarchical Reflection for Autonomous Mobile Operation
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs