MyPCBench
A benchmark for personally intelligent computer-use agents.
MyPCBench is an open-source reproducible Linux-desktop benchmark and agent-harness environment for evaluating personally intelligent computer-use agents. It provides a QEMU/KVM Ubuntu VM seeded with one coherent persona across 17 pre-logged-in simulated web applications, 184 audited tasks with rubrics, and offline LLM-as-judge grading.
Recent stories
0 linked stories
No linked stories yet.