OSWorld-Persona Benchmark Evaluates When AI Agents Should Proactively Help
Researchers introduce OSWorld-Persona, a benchmark built on OSWorld 2.0 with 108 screen recordings converted into personalized support examples, evaluating when agents should proactively offer help; deep personalization improves timing judgment by 2.4-3.0x.
2026-11-16 ~ 2026-11-16 · 2 related posts
- OSWorld-Persona: a 3,259-label benchmark for evaluating when agents should proactively help — EchoShao8899 · 2026-11-16
- OSWorld-Persona benchmark: deep personalization boosts proactive-help timing macro-F1 by 2.4-3.0x — EchoShao8899 · 2026-11-16