List your company

Vulnerability Detection on Unseen CWEs

51Affected Files CountBASE LLM
19.827.93644.1Jun 2, 2026

Evaluation results

MethodLinks
BASE LLMBase Model=CodeLlama-7B2026.065164
SFTBase Model=Qwen2.5-Coder-3B2026.064656
SFTBase Model=CodeLlama-7B2026.064450
SELF-PLAYBase Model=Qwen2.5-Coder-3B2026.064046
SFTBase Model=Qwen2.5-Coder-7B2026.063958
SELF-PLAYBase Model=Qwen2.5-Coder-7B2026.063441
SELF-PLAYBase Model=CodeLlama-7B2026.063242
BASE LLMBase Model=Qwen2.5-Coder-3B2026.062937
TSP (OURS)Base Model=CodeLlama-7B2026.062734
TSP (OURS)Base Model=Qwen2.5-Coder-3B2026.062732
BASE LLMBase Model=Qwen2.5-Coder-7B2026.062630
TSP (OURS)Base Model=Qwen2.5-Coder-7B2026.062126