Software Engineering on SWE-Skills-Bench
98Baseline AccuracyClaude Sonnet 4.5
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Claude Sonnet 4.5Baseline pass-rate stratum=Ceiling (≥ 0.9), n=37, r=12026.05 | 98 | -0.5 | 1.64 | 531 | 415 | 14.4 | 11.2 | |
| Claude Sonnet 4.5Baseline pass-rate stratum=All, n=49, r=12026.05 | 89 | 0.3 | 1.91 | 696 | 522 | 14.2 | 10.7 | |
| Claude Sonnet 4.5Baseline pass-rate stratum=Mid (0.5−0.9), n=10, r=12026.05 | 69 | 3.6 | 2.77 | 141 | 92 | 14.1 | 9.2 | |
| Claude Sonnet 4.5Baseline pass-rate stratum=Floor (< 0.5), n=2, r=12026.05 | 24 | 0 | 2.6 | 24 | 15 | 12 | 7.5 |