Skill-based Task Execution on SkillsBench
68.4AccuracyMUSE-Autoskill
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MUSE-AutoskillSkill Condition=With Human Skills, Backbone=GPT-5.5, Number of runs=52026.05 | 68.4 | 15.21 | |
| CodexSkill Condition=With Human Skills, Backbone=GPT-5.5, Number of runs=52026.05 | 67.28 | 15.17 | |
| HermesSkill Condition=With Human Skills, Backbone=GPT-5.5, Number of runs=52026.05 | 61.21 | 13.33 | |
| MUSE-AutoskillSkill Condition=Without Skills, Backbone=GPT-5.5, Number of runs=52026.05 | 53.19 | — | |
| CodexSkill Condition=Without Skills, Backbone=GPT-5.5, Number of runs=52026.05 | 52.11 | — | |
| HermesSkill Condition=Without Skills, Backbone=GPT-5.5, Number of runs=52026.05 | 47.89 | — |