Autonomous Web Navigation on VisualWebArena (test)
27Success Rate (Classifieds)CERonline
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| CERonlineMode=Online2025.06 | 27 | 38.1 | 24.4 | 31.9 | |
| CERmode=online2025.06 | 27 | 38.1 | 24.4 | 31.9 | |
| Tree SearchModel=GPT-4o, Sampling=20 times, Max Steps=52025.06 | 26.5 | 29 | 20.5 | 26.4 | |
| Tree Searchbackbone=GPT-4o, sampling per step=20, max steps=52025.06 | 26.5 | 29 | 20.5 | 26.4 | |
| BrowserGymModel=GPT-4o2025.06 | 26.2 | 28.2 | 21.1 | 26.2 | |
| BrowserGymreproduced=true2025.06 | 26.2 | 28.2 | 21.1 | 26.2 | |
| VisualWebArena2025.06 | 18.4 | 20 | 17.1 | 18.9 | |
| VisualWebArenaSet-of-Marks (SoM)=true2025.06 | 18.4 | 20 | 17.1 | 18.9 |