The Evolution and Efficacy of Open‑Source Large Language Models in Financial Risk Management (2020–2025): A PRISMA‑Guided Systematic Review

This preregistration describes a PRISMA‑2020 compliant systematic review examining how open‑source large language models (LLMs) have been developed and evaluated for financial risk management between 2020 and 2025. The review will focus on models with publicly available weights (e.g., LLaMA, Mistral, Falcon, FinGPT, and domain‑specific FinLLMs) and empirical applications in credit risk, fraud and AML, market and liquidity risk, systemic risk, and regulatory/compliance tasks. It specifies research questions, eligibility criteria, search strategy, screening procedures, data extraction, and a narrative synthesis plan