Task-free exploration on Target
77.3SR (%)AMZ → Target → WA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AMZ → Target → WATraining Setting=Sequential Curriculum, Iterations=50 + 50 + 502025.10 | 77.3 | 24.3 | |
| TargetTraining Setting=Single-Domain Specialists, Iterations=502025.10 | 77 | 52.1 | |
| WA → Target → AMZTraining Setting=Sequential Curriculum, Iterations=50 + 50 + 502025.10 | 76.8 | 23.3 | |
| AMZ + Target + WATraining Setting=Self-guided Exploration, Iterations=1502025.10 | 75.2 | 19.4 | |
| SkillWeaver*Training Setting=Sequential Curriculum, Iterations=1502025.10 | 74.2 | 18.3 | |
| Target → AMZ → WATraining Setting=Sequential Curriculum, Iterations=50 + 50 + 502025.10 | 69.2 | 18.9 | |
| AMZ → WATraining Setting=Sequential Curriculum, Iterations=75 + 752025.10 | 62.5 | 3.1 | |
| AMZTraining Setting=Single-Domain Specialists, Iterations=502025.10 | 61.5 | 3 | |
| WATraining Setting=Single-Domain Specialists, Iterations=502025.10 | 61.2 | 2.8 | |
| BaselineIterations=–2025.10 | 60.5 | — |