Distressed Real Estate Insolvency Audit Engine with LlamaIndex
Following high-profile corporate restructuring cases like China Evergrande, legal and financial institutions in Singapore and Asia require automated intelligence frameworks to audit distressed real estate holdings, untangle complex debt structures, and track off-balance-sheet guarantees across hundreds of legal filings. In this challenge, you will construct an insolvency legal discovery agent using LlamaIndex. The agent indexes thousands of court filings, corporate restructuring notices, and debt covenant documents using advanced RAG (Retrieval-Augmented Generation) with hierarchical index structures and auto-retrieval routing.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an insolvency document discovery agent using LlamaIndex advanced RAG for multi-jurisdictional distressed real estate audits.
How work is evaluated
Evaluates retrieval precision of debt obligations and accuracy of structured insolvency summaries.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
value_extraction_accuracy
Extracted offshore debt value must match ground truth target.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
citation_precision
Percentage of returned citations corresponding to target text nodes • target: 0.9 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Implement LlamaIndex hierarchical Document Summary and Vector Index engines
Configure Auto-Merging Retriever and Sentence Window Retrieval for complex legal text
Construct sub-question query engines to unearth nested guarantee structures across filings
Generate structured debt exposure matrices for creditor committees
Reference links and supporting material
Indexed legal discovery corpus comprising court restructuring orders, debt prospectus filings, and liquidator reports.
Challenge setup
Evaluates retrieval precision of debt obligations and accuracy of structured insolvency summaries.
Challenge input
JSON containing query string and target document collection ID
LlamaIndex
Required data framework for advanced sub-question indexing and retrieval.
Evaluated output
JSON containing aggregate dollar exposure, primary creditor list, and source citations
- Extracted offshore debt value must match ground truth target.
- Percentage of returned citations corresponding to target text nodes • target: 0.9 • range: 0-1
- Benchmark: LegalBench-RAG
- Citation Precision target: 0.9
- 1 public reference case
- 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
- Llama Index · Required
- SuperAGI · Optional
- Libretto · Optional
Evaluation contract
- value_extraction_accuracy · Weight 1
- citation_precision · Weight 1
Metrics
- citation_precision · Target 0.9
Benchmarks
- LegalBench-RAG
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing query string and target document collection ID
Output: JSON containing aggregate dollar exposure, primary creditor list, and source citations
Execution harness: Python
Evaluation tools
- LlamaIndex · RAG Framework
Recipe freeze marker
No recipe freeze time is recorded.