Didact: A Cross-Domain Capability Discovery System for Defence

2026-06-03 · Source: cs.CL updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, long

Summary

Didact is a prototype cross-domain capability discovery system designed for defence and defence-aligned sectors, developed as an academia-industry project for Australia. It integrates publicly available Australian defence reports and policy documents with a purpose-built knowledge graph derived from Australian research publications. The system provides natural language conversations for policy-oriented workflows, leveraging a composite Retrieval-Augmented Generation (RAG) pipeline. A key feature is its interactive Evidence Rail, which visualizes retrieved evidence and source relationships, enabling users to trace generated summaries back to underlying data. Didact maintains source isolation through separate vector stores for different access levels, such as "public" and "classified" documents, and uses structured graph retrieval for academic metadata. Evaluation shows strong retrieval quality, with a Hit@1 of 0.680 and a mean LLM response latency of 2.17 seconds, demonstrating its utility and adaptability for fragmented knowledge domains.

Key takeaway

For Directors of AI/ML deploying conversational systems in high-stakes, fragmented domains like defence, you should prioritize composite RAG architectures that integrate diverse data sources. Implement interactive evidence rails to ensure auditability and user trust, allowing traceability from generated answers to original sources. Your system must also maintain strict source isolation, partitioning data by access level to meet security requirements. This approach enhances situational awareness and supports evidence-based decision-making.

Key insights

Didact demonstrates a multi-source composite RAG system for cross-domain capability discovery with interactive evidence traceability.

Principles

• Cross-domain RAG requires multi-source synthesis.
• Evidence-grounded synthesis needs source-level control.
• Interactive evidence rails enhance traceability.

Method

Didact uses an Enricher for query priming, then a LangGraph Orchestrator for composite RAG, combining tiered document and structured graph retrieval, finally assembling an interactive Evidence Rail.

In practice

• Integrate public reports with research knowledge graphs.
• Use vector stores for documents, Neo4j for graph data.

Topics

• Retrieval-Augmented Generation
• Knowledge Graphs
• Defence Intelligence
• Cross-Domain Information Retrieval
• Evidence Traceability
• Large Language Models

Code references

• langchain-ai/langgraph

Best for: AI Architect, Research Scientist, AI Scientist, Policy Maker, Director of AI/ML

Related on AIssential

• Didact: A Cross-Domain Capability Discovery System for Defence — Didact enables cross-domain defence capability discovery by integrating fragmented knowledge sources via RAG and a knowledge graph.
• Beyond the Blank Stare: Injecting Medical Expertise into LLMs Through RAG-Powered Knowledge — RAG systems enhance LLM reliability in medicine by grounding responses in verifiable, external knowledge bases.
• Why Retrieval-Augmented Generation Fails: A Graph Perspective — RAG failures stem from shallow, fragmented internal information flow, not just poor retrieval.
• Building Multi-Tenant RAG System Architecture — Step 3: Guardrails, MCP, and Retrieval-Augmented… — Robust multi-tenant RAG systems require guardrails, reranking, and asynchronous processing for safety, accuracy, and scalability.
• RAND’s report shows that RAG, GraphRAG, and long-context AI systems can appear grounded in trusted documents while still misreading nuance, caveats, evidence strength, and partial truths. — Grounded AI systems can misinterpret nuanced information, even with citations, requiring rigorous, domain-specific evaluation.
• Don't Retrieve, Navigate: Distilling Enterprise Knowledge into Navigable Agent Skills for QA and RAG — Corpus2Skill distills document corpora into navigable hierarchical skills, enabling LLM agents to actively explore and combine evidence.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CL updates on arXiv.org.