List your company

Interactive Task Completion on ALFWorld 134 scenes (OOD)

93.28SC@7REUSERL-SegCost
-2.181622.601747.38572.1683May 29, 2026

Evaluation results

MethodLinks
REUSERL-SegCostBuffer=Global success buffer2026.0593.28
Pure Round-Length2026.0591.79
REUSERL-SegCost (no buffer)Buffer=None2026.0589.55
Vanilla GRPO2026.0579.85
Vanilla2026.051.49