ZeroHour
Product

AndroidWorld

2 mentions in 7 days · 2 in 30 days · 2 total · first seen · last

Timeline

Reflect, Revise, Reuse: Training-Free Skill Evolution for GUI Agents

EvoSkill-GUI introduces training-free skill revision for GUI agents, boosting MobileWorld, AndroidWorld, and OSWorld success by up to 16.2%.

EvoSkill-GUI is a training-free framework in which each GUI-agent skill is a structured multi-file package containing retrieval metadata, executable plans, backup localization, failure-recovery rules, and failure cases. A reflect-revise-reuse loop lets the executor make in-rollout revisions while an isolated critic diagnoses failed trajectories under strict information isolation. On MobileWorld, AndroidWorld, and OSWorld it improves multiple base models by up to +16.2%, +6.0%, and +10.5% respectively, with evolved skill libraries transferring to related tasks. Code is released on GitHub.

Hugging Face daily papers · 3d agoAI research

Google stole open source code without crediting the authors (Artemis/Minitap)

Minitap alleges Google's Artemis mobile-agent project reused its open-source mobile-use code and stripped author attribution, despite Apache 2.0 requirements.

Minitap says Google's Artemis project for automating mobile devices contains code identical to its open-source mobile-use agent, including the Hopper agent's verbatim instructions and a WhatsApp messaging example, and that a package file listing authors Pierre-Louis Favreau, Jean-Pierre Lo, and Nicolas Dehandschoewercker was replaced via an August force push removing their names. The company argues this conflicts with Apache 2.0's requirement to preserve copyright and attribution notices. Minitap also claims the AndroidWorld leaderboard ignored its later 94.8% and 100% submissions while showing Artemis at 99.1% and mobile-use at 91.4%. It has published a public factual record with archived file comparisons.

Appears with

Entities are extracted by the model from each article. Watching an entity keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.