Interactive Agent Task on ALFWorld unseen
99.3Success Rate (SR)SKILL-DISCO + CodeAct
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SKILL-DISCO + CodeActBackbone=GPT-4o2026.06 | 99.3 | 3.2 | |
| CodeActBackbone=GPT-4o2026.06 | 96.3 | 3.6 | |
| SKILL-DISCO + ReActBackbone=GPT-4o2026.06 | 92.4 | 8.8 | |
| ReActBackbone=GPT-4o2026.06 | 82 | 19.3 | |
| AWMofflineBackbone=GPT-4o2026.06 | 54.5 | 11.3 | |
| ASIofflineBackbone=GPT-4o2026.06 | 47 | 11.4 |