DefenceRAG: Procurement & Policy Reasoning Challenge
Build intelligent Retrieval-Augmented Generation (RAG) systems to understand and reason over complex defence policy documents. Participants are required to develop models that can interpret defence procurement manuals, financial delegation rules, and naval regulations to answer domain-specific questions with accurate reasoning and grounded references.
Unlike traditional QA systems, this challenge emphasizes: - Multi-document reasoning - Policy interpretation - Explainable answers with citations
The goal is to simulate real-world decision-making scenarios in defence administration using AI. Top-performing participants may be considered for internships and collaboration opportunities on INICAI-driven projects.
Understanding defence procurement and policy documents requires more than simple information retrieval. These documents are highly structured, rule-driven, and often involve layered decision-making based on financial limits, authority hierarchies, and procedural constraints. Interpreting them correctly is essential for making compliant and informed decisions in real-world defence systems.
In this competition, participants are challenged to build RAG systems that can read, interpret, and reason over official defence documents, including: - Defence Procurement Manual (DPM) - Delegation of Financial Powers for Defence Services (DFPDS) - Navy Regulations (Part 1–4)
These documents define how procurement processes are executed, who holds financial authority, and how operations are governed across different levels of command. Unlike traditional question-answering tasks, this problem requires systems to go beyond extracting text. Models must identify relevant sections, understand policy constraints, and generate answers that are logically consistent with the rules described in the documents. The ability to combine information across multiple sources and produce clear, explainable reasoning is central to this challenge.
Participants will work on tasks such as identifying approval authorities, interpreting procurement procedures, and validating compliance with policy guidelines. A strong solution should not only produce accurate answers but also justify them using grounded references from the provided documents.