Downstream task execution on SkillsBench
29.26Reward Mean (%)SKILLRAE
Evaluation Results
| Method | Links | |
|---|---|---|
| SKILLRAE2026.05 | 29.26 | |
| Benchmark-native Curated Skills2026.05 | 26.2 | |
| SkillRouter2026.05 | 22.04 | |
| Vanilla Retrieval2026.05 | 19.44 | |
| AgentSkillOS2026.05 | 17.24 | |
| LLM-based Retrieval2026.05 | 16.34 |