List your company

Software Engineering Skill Evaluation on SWE-Skills-Bench implementing-agent-modes

30Pass RateFull Skill Text
19.622.32527.7Jun 15, 2026

Evaluation results

MethodLinks
Full Skill TextSetting=Full Skill Text prompting2026.063088.530.11
S2LSetting=Our method2026.063021.910.46
Vanilla LLMSetting=Vanilla baseline2026.0620——