Lifelong Agent Interaction on AlfWorld
71.28Success RatePROACTAGENT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PROACTAGENTRegime=Ours2026.04 | 71.28 | 12.73 | |
| GRPO (online)Regime=Online baselines2026.04 | 68.71 | 13.99 | |
| GRPORegime=Offline baselines2026.04 | 67.69 | 17.84 | |
| GRPO+ReflexionRegime=Online baselines2026.04 | 67.18 | 16.42 | |
| SFTRegime=Offline baselines2026.04 | 54.36 | 26.78 | |
| ReflexionRegime=Online baselines2026.04 | 36.92 | 38.57 | |
| Mem0Regime=Online baselines2026.04 | 33.85 | 32.11 | |
| MemoryBankRegime=Online baselines2026.04 | 32.31 | 37.14 | |
| AWMRegime=Online baselines2026.04 | 31.28 | 33.56 | |
| BaselineRegime=Offline baselines2026.04 | 27.69 | 32.85 |