Sub-task Completion on AutoPenBench
212AC (Count)Qwen3-32B-finetune
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Qwen3-32B-finetuneFramework=xOffense, Note=Ours2025.09 | 212 | 173 | 71 | 176 | 334 | 966 | |
| Llama3.1-405BFramework=VulnBot2025.09 | 107 | 116 | 40 | 75 | 186 | 524 | |
| Llama3.3-70BFramework=VulnBot2025.09 | 87 | 106 | 41 | 65 | 166 | 465 | |
| Llama3.1-405BFramework=Base2025.09 | 61 | 66 | 22 | 44 | 67 | 260 | |
| Qwen3-32BFramework=Base2025.09 | 60 | 70 | 10 | 70 | 155 | 365 | |
| Llama3.3-70BFramework=Base2025.09 | 46 | 83 | 36 | 68 | 99 | 332 | |
| Llama3.1-405BFramework=PentestGPT2025.09 | 27 | 40 | 15 | 43 | 56 | 181 |