Source record

Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2608.16921.

Pipeline notes

The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.

A multi-agent retrieval-augmented generation framework (MITRE-SAGE) for cybersecurity question-answering that decomposes tasks into query interpretation, evidence retrieval, and answer synthesis. A new 3,000-pair benchmark (MITRE-QA) is introduced, and a lightweight Qwen2.5-7B/14B configuration outperforms standalone LLMs and conventional RAG on five of eight tasks.

Concrete multi-agent RAG architecture with stated model sizes and a new evaluation benchmark, serving as a reference design for domain-specific agentic QA pipelines.

Cited by

No citations on record.