| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| SWE-bench Verified | TACT-GATE | Resolution Rate73.3 | 63 | 2mo ago | |
| SWE-Bench Verified | Terminus 2 | Success Rate (SR)61.6 | 14 | 1mo ago | |
| SWE-Gym (Holdout) | DAgger-style | Task Resolution Rate19 | 12 | 2mo ago | |
| SWE-BENCH LIVE | Qwen3-Coder 480B/A35B Instruct | Resolution Rate24.7 | 11 | 5mo ago | |
| SWE-bench In-Distribution | SkillFlow | Resolved Rate52.34 | 9 | 2mo ago | |
| SWE-agent task distribution | Governor | Tokens per Success74,000 | 3 | 1mo ago |