Sub-workflow Execution on OS-Marathon Expense Report - Website System Environment
95SWAHuman
Evaluation Results
| Method | Links | |
|---|---|---|
| HumanStep=1502026.01 | 95 | |
| HumanStep=2002026.01 | 95 | |
| HumanStep=1002026.01 | 70 | |
| AgentS2.5 + GPT5 w/ FCWDStep=1502026.01 | 37.5 | |
| AgentS2.5 + GPT5 w/ FCWDStep=2002026.01 | 37.5 | |
| HumanStep=502026.01 | 30 | |
| AgentS2.5 + GPT5 w/ FCWDStep=1002026.01 | 30 | |
| AgentS2.5 + GPT5 w/ FCWDStep=502026.01 | 12.5 | |
| AgentS2.5 + GPT5Step=1002026.01 | 5 | |
| AgentS2.5 + GPT5Step=1502026.01 | 5 | |
| AgentS2.5 + GPT5Step=2002026.01 | 5 | |
| OpenCUA-7BStep=502026.01 | 0 | |
| OpenCUA-7BStep=1002026.01 | 0 | |
| OpenCUA-7BStep=1502026.01 | 0 | |
| OpenCUA-7BStep=2002026.01 | 0 | |
| UI-TARS-1.5-7BStep=502026.01 | 0 | |
| UI-TARS-1.5-7BStep=1002026.01 | 0 | |
| UI-TARS-1.5-7BStep=1502026.01 | 0 | |
| UI-TARS-1.5-7BStep=2002026.01 | 0 | |
| AgentS2.5 + GPT5Step=502026.01 | 0 |