Agentic Task Accuracy on ALFWorld
79.1Task AccuracyHard final
Evaluation Results
| Method | Links | |
|---|---|---|
| Hard finalSkill Setting=Hard final2026.06 | 79.1 | |
| SOFTSKILLSkill Setting=Val-selected, Prefix Length=auto2026.06 | 71.6 | |
| Oracle grid bestSkill Setting=Oracle grid best, Prefix Length=auto2026.06 | 71.6 | |
| Hard initialSkill Setting=Hard initial2026.06 | 64.9 | |
| No skillSkill Setting=No skill2026.06 | 57.5 |