// AI news · 2026-11-02 · Apple Machine Learning Research
Apple ML Research Proposes Online RL Framework for iOS UI Navigation
Apple ML Research proposes an online reinforcement learning framework for iOS UI navigation, decoupling agents into planning and grounding components.
Published 2026-11-02 08:00 (Beijing time)Source Apple Machine Learning ResearchAuto-summarised, fact-checked against source
01Summary
It uses LLM-based reward signals with step-level feedback and terminal rewards, enhanced by GiGPPO for value estimation stability. Evaluations show significant outperformance over SFT and terminal-only baselines, especially in out-of-distribution generalization to Android environments.
Summarised from the original article by an open-source model, then checked automatically: every number and English name must appear in the source. The original article is authoritative.
02Key points
- Decouples agent into planning and grounding components
- Uses LLM-based reward with step-level and terminal feedback
- Introduces GiGPPO for improved value estimation via visual similarity
- Outperforms SFT baselines in Android out-of-distribution generalization
03Source
Online Dense Reinforcement Learning with Planning-Grounding Agent for UI Navigation on Live iOS Apps
Apple Machine Learning Research · published 2026-11-02 08:00 Beijing time
05Related news
From AIMap AI news: collected from 1,100+ official and media sources. Summaries only restate the original article; follow the link for the full story. All stories of 2026-11-02 · Archive by date · All AI news · Today