A benchmark called iPrefBench has been introduced to evaluate personalized mobile-use agents on underspecified real-world user requests, according to the benchmark's authors. The benchmark targets a gap the authors identify: computer-use agents can execute fully specified tasks across apps, but real-world requests are often underspecified, so agents must infer user preferences to act correctly. The work matters because it shifts agent evaluation toward user-centric personalization rather than task completion alone.
An agent that truly knows and acts on behalf of its user has been a long-standing promise of AI assistants. Computer-use agents bring this promise within reach by capably executing fully specified tasks across apps. However, real-world user requests are often underspecified. To act correctly for the
EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.