arXiv q-fin· Masoud Soleimani·· 5 hr agoAI Score62
arXiv Paper: Generating LLM Counterfactual Stress Scenarios with a Hybrid Prompt-RAG Pipeline for Portfolio Risk Simulation
LLM-Generated Counterfactual Stress Scenarios for Portfolio Risk Simulation via Hybrid Prompt-RAG Pipeline
AI summary
Masoud Soleimani released a 30-page arXiv paper proposing an auditable large language model macro-financial stress testing pipeline. This pipeline generates G7 macro scenarios including GDP growth, inflation, and policy rates through structured prompts and optional retrieval of country fundamentals and news, converts them into deviations from the IMF World Economic Outlook baseline, and then maps them to portfolio losses via a three-channel PCA factor model, with the macro-to-factor mapping pre-specified rather than estimated.
Source:arXiv q-fin · arxiv.org
Academic papers#Papers / Research#Risk Management#Large Language Model Applications#Portfolio Construction#arXiv