Embodied Task Completion on ALFWorld (Unseen tasks)
100Look Success RateEvolving-RL
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Evolving-RLSkill Injection=false2026.05 | 100 | 61.2 | 81.1 | |
| Evolving-RLSkill Injection=true2026.05 | 95.6 | 81.2 | 88.6 | |
| SkillRLSkill Injection=false2026.05 | 81.2 | 55.6 | 68.8 | |
| ExpeLSkill Injection=false2026.05 | 67 | 6 | 37.4 | |
| GRPOSkill Injection=true2026.05 | 61.1 | 27.1 | 44.6 | |
| GRPOSkill Injection=false2026.05 | 53.3 | 12.9 | 33.7 | |
| Base ModelSkill Injection=false2026.05 | 18.9 | 36.5 | 27.4 | |
| Base ModelSkill Injection=true2026.05 | 16.7 | 36.5 | 26.3 | |
| ReasoningBankSkill Injection=false2026.05 | 11.1 | 35.3 | 22.9 | |
| MementoSkill Injection=false2026.05 | 0 | 41.2 | 20 |