Trajectory Verification on SWE-bench Verified Qwen3-Coder-Max trajectories
74.6RM@32SWE-RM-30A3B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SWE-RM-30A3BType=EF (execution-free)2025.12 | 74.6 | 0.768 | 0.047 | |
| DEEP SWEType=EB (execution-based)2025.12 | 67.6 | — | — | |
| DEEP SWEType=EF (execution-free)2025.12 | 66.2 | 0.74 | 0.139 | |
| SWE-GYMType=EF (execution-free)2025.12 | 65.4 | 0.752 | 0.283 | |
| AGENTLESSType=EB (execution-based)2025.12 | 65 | — | — |