SkillForge: Self-Distilling Agents for Project-Specific Issue Resolution

Open paper
Benchmark Model Rank Results
code-generation-on-swe-bench-proSkillForge (GPT-5-mini) (extra data)#30Resolved: 51.7
code-generation-on-swe-bench-verifiedSkillForge (DeepSeek-V3.2) (Mini-SWE-Agent) (extra data)#33Accuracy: 72.2

Rank counts only results with a code link.