Vulnerability Detection on PrimeVul
67.93F1 ScoreQwen2.5-32B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen2.5-32BApproach=CPRVUL2026.02 | 67.93 | 67.78 | 67.62 | 68.23 | |
| Qwen2.5-7BApproach=CPRVUL2026.02 | 67.19 | 62.7 | 59.97 | 76.4 | |
| Gemini-2.5-FlashSource Type=Closed-Source2026.02 | 64.16 | — | 51.76 | 84.37 | |
| GPT-4.1Approach=Baseline2026.02 | 63.39 | 51.9 | 51.17 | 83.3 | |
| SecCoderX RM 8BSource Type=Open-Source, Model Scale=8B2026.02 | 61.96 | — | 50.29 | 80.69 | |
| UniXcoderApproach=Baseline2026.02 | 61.55 | 56.65 | 55.34 | 69.45 | |
| GPT-4.1Source Type=Closed-Source2026.02 | 59.9 | — | 54.53 | 66.44 | |
| R2Vul 7BSource Type=Open-Source, Model Scale=7B2026.02 | 59.65 | — | 49.54 | 74.94 | |
| VulAgentFramework Type=Multi-Agent Frameworks, Dataset Partition=PrimeVul2026.02 | 56.2 | 58.4 | — | — | |
| SecCodePRMParameters=7B2026.02 | 55.71 | 92.57 | — | — | |
| CodeBERTApproach=Baseline2026.02 | 55.62 | 53.21 | 52.78 | 63.03 | |
| Qwen3-14BSource Type=Open-Source, Model Scale=14B2026.02 | 44.75 | — | 55.03 | 37.7 | |
| Qwen3-8BSource Type=Open-Source, Model Scale=8B2026.02 | 41.35 | — | 50.16 | 35.17 | |
| Qwen2.5-Coder-14B-InstSource Type=Open-Source, Model Scale=14B, Instruction Tuned=true2026.02 | 24.36 | — | 47.97 | 16.32 | |
| Llama4-17BParameters=17B2026.02 | 21.43 | 85.03 | — | — | |
| GPT-oss-120B-hParameters=120B, Variant=h2026.02 | 16.52 | 64.29 | — | — | |
| QC2.5-7BParameters=7B2026.02 | 13.51 | 82.26 | — | — | |
| GPT-oss-120B-mParameters=120B, Variant=m2026.02 | 12.53 | 71.04 | — | — | |
| GPT-oss-120B-lParameters=120B, Variant=l2026.02 | 9.79 | 80.18 | — | — | |
| Codestral-22BParameters=22B2026.02 | 9.3 | 70.31 | — | — | |
| QC2.5-32BParameters=32B2026.02 | 6.15 | 89.29 | — | — | |
| QC3-30BParameters=30B2026.02 | 4 | 88.55 | — | — | |
| Gemma3-12BParameters=12B2026.02 | 2.4 | 80.63 | — | — | |
| Qwen2.5-Coder-7B-InstSource Type=Open-Source, Model Scale=7B, Instruction Tuned=true2026.02 | 0 | — | 0 | 0 |