Skip to main content
Back to challenges
Challenge

MAS Regulatory Intelligence & Asset Screening Agent with LlamaIndex

Asset management firms in Singapore face strict compliance verification under updated Monetary Authority of Singapore (MAS) asset hub guidelines. Build an intelligent regulatory screening RAG pipeline using LlamaIndex and Gemini 2.5 Flash to index regulatory circulars and evaluate fund structure compliance with 96% citation accuracy.

CybersecurityHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Develop a high-performance Retrieval-Augmented Generation system using LlamaIndex and Gemini 2.5 Flash to automatically index MAS regulatory PDF publications and query structural compliance conditions.

Delivery guide

How work is evaluated

Evaluation

Evaluates context recall and accuracy of regulatory citation answers.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

Citation Verifiability

Ensure every assertion in the generated answer includes a matching source document citation.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

faithfulness

RAG answer faithfulness score evaluated by LlamaIndex Evaluator • target: 0.96 • range: 0.9-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Build chunking and vector index pipelines for regulatory publications with LlamaIndex

  • Integrate Gemini 2.5 Flash for ultra-fast, long-context regulatory retrieval

  • Implement citation verification logic to prevent hallucinated compliance guidance

Resources and assets

Reference links and supporting material

Dataset notes

Collection of 50 MAS regulatory PDFs and regulatory guidance papers published in 2026.

Challenge setup

Evaluates context recall and accuracy of regulatory citation answers.

Challenge input

Natural language query regarding MAS asset management policy

LlamaIndex

Assigned RAG and data framework.

Gemini 2.5 Flash

Assigned generative model for fast context retrieval.

Evaluated output

JSON response with compliant status, answer text, and target document citations

Evaluation checks
  • Ensure every assertion in the generated answer includes a matching source document citation.
  • RAG answer faithfulness score evaluated by LlamaIndex Evaluator • target: 0.96 • range: 0.9-1
Configured evaluation evidence
  • Benchmark: RegRAG-Bench
  • Faithfulness target: 0.96
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • Llama Index · Required
  • LlamaIndex · Optional
  • Llama 3.2 · Optional

Evaluation contract

  • Citation Verifiability · Weight 1
  • faithfulness · Weight 1

Metrics

  • faithfulness · Target 0.96

Benchmarks

  • RegRAG-Bench
  • security

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: Natural language query regarding MAS asset management policy

    Output: JSON response with compliant status, answer text, and target document citations

    Execution harness: Python

Evaluation tools

  • LlamaIndex · RAG Framework
  • Gemini 2.5 Flash · Language Model

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about MAS Regulatory Intelligence & Asset Screening Agent with LlamaIndex