Risk Score Aggregation And Policy Based Decision Making

via “risk management multi-agent assessment with portfolio approval”

TradingAgents: Multi-Agents LLM Financial Trading Framework

Unique: Implements a three-agent risk assessment team (VaR, Correlation, Liquidity) that independently evaluates trades, with a Portfolio Manager agent that synthesizes their outputs and has final veto authority. Each risk agent uses deep thinking LLM to reason about risk dimensions, rather than using simple rule-based checks, enabling nuanced risk assessment that accounts for market context.

vs others: More comprehensive than single-metric risk checks (e.g., VaR-only) because it evaluates multiple risk dimensions independently and synthesizes them. More explainable than black-box risk models because each agent produces reasoning traces that justify approval/rejection decisions, useful for compliance and audit trails.