Vulnerability Detection on Unseen CWEs
51Affected Files CountBASE LLM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BASE LLMBase Model=CodeLlama-7B2026.06 | 51 | 64 | |
| SFTBase Model=Qwen2.5-Coder-3B2026.06 | 46 | 56 | |
| SFTBase Model=CodeLlama-7B2026.06 | 44 | 50 | |
| SELF-PLAYBase Model=Qwen2.5-Coder-3B2026.06 | 40 | 46 | |
| SFTBase Model=Qwen2.5-Coder-7B2026.06 | 39 | 58 | |
| SELF-PLAYBase Model=Qwen2.5-Coder-7B2026.06 | 34 | 41 | |
| SELF-PLAYBase Model=CodeLlama-7B2026.06 | 32 | 42 | |
| BASE LLMBase Model=Qwen2.5-Coder-3B2026.06 | 29 | 37 | |
| TSP (OURS)Base Model=CodeLlama-7B2026.06 | 27 | 34 | |
| TSP (OURS)Base Model=Qwen2.5-Coder-3B2026.06 | 27 | 32 | |
| BASE LLMBase Model=Qwen2.5-Coder-7B2026.06 | 26 | 30 | |
| TSP (OURS)Base Model=Qwen2.5-Coder-7B2026.06 | 21 | 26 |