Vulnerability Injection on CWEval Python
66.67Attack Success Rate (ASR)XOXO + GCGS
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| XOXO + GCGSModel=DeepSeek Coder 6.7B2025.03 | 66.67 | 2,984 | |
| XOXOModel=DeepSeek Coder 33B2025.03 | 66.67 | 3,143 | |
| XOXOModel=DeepSeek Coder 6.7B2025.03 | 64.44 | 3,128 | |
| XOXO + GCGSModel=DeepSeek Coder 33B2025.03 | 63.97 | 3,239 | |
| XOXO + GCGSModel=Codestral 22B2025.03 | 62.58 | 2,927 | |
| XOXOModel=Codestral 22B2025.03 | 60.3 | 3,077 | |
| XOXO + GCGSModel=Qwen 2.5 Coder 7B2025.03 | 55 | 3,813 | |
| XOXO + GCGSModel=Llama 3.1 8B2025.03 | 54 | 3,719 | |
| XOXO + GCGSModel=GPT 4o2025.03 | 50 | 4,144 | |
| XOXOModel=Llama 3.1 8B2025.03 | 48.89 | 4,059 | |
| XOXOModel=Qwen 2.5 Coder 7B2025.03 | 48.33 | 3,962 | |
| XOXOModel=Claude 3.5 Sonnet v22025.03 | 40 | 4,690 | |
| XOXO + GCGSModel=Qwen 2.5 Coder 32B2025.03 | 27.69 | 5,839 | |
| XOXOModel=Qwen 2.5 Coder 32B2025.03 | 23.08 | 5,927 |