Evaluation Tasks
Tasks with user goals, app context, and success criteria to test your agent.
Explore ↗We provide evaluation data and RL environments to improve personal agents across apps. Tasks are designed using real user insights.
01/03Users
Tasks and constraints built from real user insights and synthetic personas.
02/03Evaluate
Check task completion, preferences, and permissions.
Dining · Travel · Shopping · Money · Health · Home · Work · and more
03/03Train
Digital worlds, user simulators, and rewards for intermediate steps and final outcomes.
Can personal agents complete everyday tasks?
A final exam for everyday personal agents.
Code that meets the user’s requirements.
Using apps and websites for different users.
Helpful answers for different user needs.
Science tasks for different user backgrounds.
Following user preferences and constraints.