Interactive Agent Task on WebArena (evaluation)
29.1Success Rate (SR)SKILL-DISCO + ReAct
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SKILL-DISCO + ReActBackbone=GPT-4o2026.06 | 29.1 | 5.1 | |
| ASIofflineBackbone=GPT-4o2026.06 | 24.6 | 5.7 | |
| ReActBackbone=GPT-4o2026.06 | 23.9 | 5.9 | |
| SKILL-DISCO + CodeActBackbone=GPT-4o2026.06 | 22.9 | 8.5 | |
| AWMofflineBackbone=GPT-4o2026.06 | 21.2 | 5.9 | |
| CodeActBackbone=GPT-4o2026.06 | 20 | 10.9 |