Research map / Reinforcement learning / Visual and representation RL
MobileRL: Online Agentic Reinforcement Learning for Mobile GUI Agents
Yifan Xu, Xiao Liu, Xinghan Liu, Jiaqi Fu, Jiayu Huang, Hanchen Zhang, Bohao Jing, Shudan Zhang, Yuting Wang, Zhao wenyi, Yuxiao Dong
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add AdaGRPO RL stage after SFT (MobileRL-9B) (ablation) | MobileRL-9B · AndroidWorld | success rate (%): 66.2 -> 80.2 | Improved |
| add AdaGRPO RL stage after SFT (MobileRL-9B) (ablation) | MobileRL-9B · AndroidLab | success rate (%): 45.0 -> 53.6 | Improved |
| add RL stage (ablation) | MobileRL-9B · AndroidControl-High | SR (%): 77.5 -> 73.9 | Hurt |
| add RL stage (ablation) | MobileRL-7B · AndroidControl-Low | SR (%): 91.9 -> 91.4 | No clear effect |
| add RL stage (ablation) | MobileRL-7B · AndroidControl-High | grounding: 73.6 -> 71.5 | Hurt |
| add Reasoning-Free SFT before Reasoning SFT (ablation) | MobileRL · AndroidWorld | success rate (%) | Improved |
| add AdaPR replay (ablation) | MobileRL · AndroidWorld | reward / success rate | Improved |
| add SPA reward adjustment (ablation) | MobileRL · AndroidWorld | step efficiency | Improved |
| add FCF failure curriculum filtering (ablation) | MobileRL · AndroidWorld | reward | Improved |
| use original instead of compressed images (GLM-4.1V-9B) (ablation) | MobileRL-9B AdaGRPO · AndroidWorld | success rate (%): 75.8 -> 80.2 | Improved |
| use original instead of compressed images (Qwen2.5-VL) (ablation) | MobileRL-7B · AndroidWorld | success rate (%) | No clear effect |
| use original instead of compressed images (ablation) | Reasoning SFT GLM-4.1V-9B · AndroidWorld | success rate (%): 66.1 -> 66.2 | No clear effect |
Closest papers by meaning
- GUI Exploration Lab: Enhancing Screen Navigation in Agents via Multi-Turn Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- DistRL: An Asynchronous Distributed Reinforcement Learning Framework for On-Device Control Agent (ICLR 2025 · similarity 0.89)
- MobileIPL: Enhancing Mobile Agents Thinking Process via Iterative Preference Learning (ICLR 2026 · similarity 0.89)
- SE-GUI: Enhancing Visual Grounding for GUI Agents via Self-Evolutionary Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- WebRL: Training LLM Web Agents via Self-Evolving Online Curriculum Reinforcement Learning (ICLR 2025 · similarity 0.88)
- Succeed or Learn Slowly: Sample Efficient Off-Policy Reinforcement Learning for Mobile App Control (NeurIPS 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
