GUI Automation on MiniWoB++ Hard - 9 types (test)
1Token Count (k)AutoRPA (code only)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AutoRPA (code only)LLM Backbone=GPT-4.12026.05 | 1 | 80 | |
| AutoRPALLM Backbone=GPT-4.12026.05 | 1.4 | 91.1 | |
| AdaPlanner (one demo)LLM Backbone=GPT-4.1, human examples=12026.05 | 13.1 | 24.4 | |
| AdaPlannerLLM Backbone=GPT-4.12026.05 | 15.1 | 74.1 | |
| ReAct†LLM Backbone=GPT-4.12026.05 | 16.2 | 84.4 | |
| RCILLM Backbone=GPT-4.12026.05 | 19.7 | 57.8 | |
| AutoManualLLM Backbone=GPT-4.12026.05 | 23.2 | 91.1 |