Proof-of-Vulnerability test generation on 33-sample dataset
0.9697C Success RatePoVSmithM
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| PoVSmithMCoding Agent=Mistral Vibe 2.0 with Devestral 22026.05 | 0.9697 | 0.8788 | 0.5455 | 97 | 0.3636 | 36 | 73 | |
| PoVSmithCoding Agent=GPT-5.2-Codex2026.05 | 0.9394 | 0.9697 | 0.9394 | 94 | 0.697 | 70 | 73 | |
| PoVSmithGCoding Agent=Gemini Code Assist with Gemini-2.5-pro2026.05 | 0.697 | 0.6364 | 0.3939 | 70 | 0.3333 | 33 | 79 | |
| Zhang et al's tool2026.05 | 0.4848 | — | — | 48 | 0.1515 | 15 | — |