MARS: Malware Analysis with Rule-Based Scoring of LLM Claims

Hyeongjun Choi

Large language models can triage malware through direct verdicts or behavioral claims scored by an external policy. We present MARS, a malware triage framework, and compare direct…