SKILLLITE: Evidence-Guided Malicious Skill Auditing with Compact LLMs
Experiments show that SKILLLITE improves malicious Skill detection across different compact LLM backbones and outperforms existing representative auditing baselines, and generalizes to behaviorally confirmed in-the-wild malicious Skills.
Hao-Ran Ou, Ge-Lei Deng, Xuan-Ye Zhang et al.
· 0 citations