List your company

GUI Task Automation on OS-Marathon Transcript domain (Level 1)

100SWAHuman
-4235077Jan 28, 2026

Evaluation results

MethodLinks
HumanStep=1002026.01100100
AgentS2.5 + GPT5 w/ FCWDStep=1002026.0191.7450
HumanStep=502026.017550
AgentS2.5 + GPT5 w/ FCWDStep=502026.0166.2925
AgentS2.5 + GPT5Step=1002026.0127.0825
AgentS2.5 + GPT5Step=502026.0119.940
OpenCUA-7BStep=502026.0100
OpenCUA-7BStep=1002026.0100
UI-TARS-1.5-7BStep=502026.0100
UI-TARS-1.5-7BStep=1002026.0100