Research & academia
How Benchmarks Mis-Score Computer-Use Agents
A paper audits how computer-use benchmarks score failed trajectories.
Why it matters: Defects in benchmark scoring can change how agent performance is interpreted; the finding is limited to the audited sample.