MITRE-SAGE framework enhances cybersecurity QA with multi-agent approach

Researchers have developed MITRE-SAGE, a multi-agent framework designed to enhance question-answering capabilities in cybersecurity. This system integrates semantic and structural cybersecurity knowledge to improve reliability and reduce hallucinations, a common issue with standard LLMs in this domain. MITRE-SAGE decomposes tasks into query interpretation, evidence retrieval, and answer synthesis, proving effective for vulnerability assessment and threat profiling. To evaluate its performance, a new benchmark called MITRE-QA was created, featuring 3,000 question-answer pairs, which MITRE-SAGE consistently outperformed against baseline methods, even with a lightweight configuration using Qwen2.5 models. AI

IMPACT This framework could significantly improve the accuracy and reliability of AI in cybersecurity operations, addressing information overload and hallucination issues.

RANK_REASON The cluster describes a new research paper detailing a novel framework and benchmark for AI in cybersecurity. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Hugging Face
• MITRE-QA
• MITRE-SAGE
• Qwen2.5-14B
• Qwen2.5-7B

• paper
• model release
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →