Code Security Evaluation on DiverseVul C/C++
142VULNS CountSelf-Play
Evaluation Results
| Method | Links | |
|---|---|---|
| Self-PlayBase Model=Qwen2.5-Coder-7B2026.06 | 142 | |
| SFTBase Model=Qwen2.5-Coder-7B2026.06 | 140 | |
| Self-PlayBase Model=Qwen2.5-Coder-3B2026.06 | 130 | |
| Base LLMBase Model=CodeLlama-7B2026.06 | 115 | |
| SFTBase Model=CodeLlama-7B2026.06 | 110 | |
| SFTBase Model=Qwen2.5-Coder-3B2026.06 | 107 | |
| Base LLMBase Model=Qwen2.5-Coder-7B2026.06 | 106 | |
| Self-PlayBase Model=CodeLlama-7B2026.06 | 103 | |
| TSP (OURS)Base Model=CodeLlama-7B2026.06 | 94 | |
| Base LLMBase Model=Qwen2.5-Coder-3B2026.06 | 93 | |
| TSP (OURS)Base Model=Qwen2.5-Coder-7B2026.06 | 57 | |
| TSP (OURS)Base Model=Qwen2.5-Coder-3B2026.06 | 41 |