Software Engineering Skill Evaluation on SWE-Skills-Bench bash-defensive-patterns
50Pass RateVanilla LLM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Vanilla LLMSetting=Vanilla baseline2026.06 | 50 | — | — | |
| S2LSetting=Our method2026.06 | 40 | -26.6 | -0.38 | |
| Full Skill TextSetting=Full Skill Text prompting2026.06 | 30 | 12.95 | -1.54 |