Software Engineering Performance on SWE-bench Lite (Accuracy)
16.1AccuracyTOOLSELF
Evaluation Results
| Method | Links | |
|---|---|---|
| TOOLSELFBase Model=Qwen3-14B2026.02 | 16.1 | |
| SWE-SearchBase Model=Qwen3-14B2026.02 | 14.6 | |
| SWE AgentBase Model=Qwen3-14B2026.02 | 14.2 | |
| Vanilla AgentBase Model=Qwen3-14B2026.02 | 13.3 | |
| TOOLSELFBase Model=Qwen3-8B2026.02 | 12.4 | |
| SWE-SearchBase Model=Qwen3-8B2026.02 | 11.6 | |
| SWE AgentBase Model=Qwen3-8B2026.02 | 10.9 | |
| Vanilla AgentBase Model=Qwen3-8B2026.02 | 10.1 |