Software Engineering Skill Evaluation on SWE-Skills-Bench (changelog-automation)
60Pass RateFull Skill Text
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Full Skill TextSetting=Full Skill Text prompting2026.06 | 60 | 7.34 | 5.45 | |
| S2LSetting=Our method2026.06 | 60 | -1.32 | 30.19 | |
| Vanilla LLMSetting=Vanilla baseline2026.06 | 20 | — | — |