CORTEX: Collaborative LLM Agents for High-Stakes Alert Triage

Capture

• arXiv: 2510.00311
• DOI: https://doi.org/10.48550/arXiv.2510.00311
• Submitted: 2025-09-30
• Source URL: https://arxiv.org/abs/2510.00311

Abstract Summary

CORTEX proposes a collaborative multi-agent LLM architecture for high-stakes alert triage. Specialized agents inspect behavior sequences, query external systems, gather evidence, and synthesize auditable decisions. The paper also claims a dataset of fine-grained SOC investigations from production environments.

Why It Matters For The Wiki

This source is valuable for multi-agent AI SOC design, auditability, evidence-gathering workflows, and comparing specialized-agent architectures against single-agent triage approaches.

Recommended Ingest Notes

• Extract claims about multi-agent division of labor, auditable decision-making, dataset provenance, and false-positive reduction.
• Connect to MAS security, agent identity, SOC provenance, and alert-triage benchmark gaps.