Interactive Task Completion on ALFWorld 140 scenes (IID)
97.14Success Count @ 7 StepsREUSERL-SegCost
Evaluation results
| Method | Links | |
|---|---|---|
| REUSERL-SegCostBuffer=Global success buffer2026.05 | 97.14 | |
| Pure Round-Length2026.05 | 96.43 | |
| REUSERL-SegCost (no buffer)Buffer=None2026.05 | 93.57 | |
| Vanilla GRPO2026.05 | 84.29 | |
| Vanilla2026.05 | 3.57 |