Embodied Agent Task Completion on ALFWorld
7.8Success Rate DeltaReasoningBank
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ReasoningBankCurator=Qwen3-8B, Executor=Qwen3-8B2026.06 | 7.8 | — | 20.1 | |
| HDSOCurator=Qwen3-8B, Executor=Qwen3.5-9B2026.06 | 7.8 | 52.1 | 20.8 | |
| HDSOCurator=Qwen3-8B, Executor=Qwen3-8B2026.06 | 6.9 | 47.4 | 20.9 | |
| SkillOS-baseCurator=Qwen3-8B, Executor=Qwen3-8B2026.06 | 5.2 | — | 20.4 | |
| HDSOCurator=Qwen3.6-27B, Executor=Qwen3.6-27B2026.06 | 4 | 65.7 | 17.3 | |
| MemPCurator=Qwen3-8B, Executor=Qwen3-8B2026.06 | 1.8 | — | 21 | |
| Executor-onlyCurator=none, Executor=Qwen3-8B2026.06 | 0 | 40.5 | 22.3 | |
| Executor-onlyCurator=none, Executor=Qwen3.5-9B2026.06 | 0 | 44.3 | 21.9 | |
| Executor-onlyCurator=none, Executor=Qwen3.6-27B2026.06 | 0 | 61.7 | 17.9 |