Skill-augmented Task Completion on SkillsBench 77 tasks (test)
0.46Mean RewardHuman
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| HumanEvaluation Protocol=Method D, Number of trials (k)=22026.06 | 0.46 | 0.098 | 100 | |
| SKILLAXEEvaluation Protocol=Method D, Number of trials (k)=22026.06 | 0.273 | 0.084 | 11.4 | |
| LLM-selfEvaluation Protocol=Method D, Number of trials (k)=22026.06 | 0.258 | 0.088 | — | |
| No-skillEvaluation Protocol=Method D, Number of trials (k)=22026.06 | 0.249 | 0.085 | — |