Agent Task Execution on AppWorld online mode
57.7AccuracySkillComposer-4B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SkillComposer-4BExecutor=Qwen3.5-4B2026.06 | 57.7 | 88.3 | |
| Qwen3.5-4BExecutor=Qwen3.5-4B, Status=untrained composer2026.06 | 57.1 | 86 | |
| VanillaExecutor=Qwen3.5-4B, Note=No Skill2026.06 | 55.2 | 88.1 |