Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FINRS: A Risk-Sensitive Trading Framework for Real Financial Markets

Bijia Liu, Ronghao Dang

TL;DR

FinRS tackles risk management in LLM-based trading by embedding risk-awareness into perception, decision making, and reward. It integrates three modules: Market Perception and Analysis with hierarchical memory, Risk-Sensitive Decision Making with dual agents and dynamic position sizing using CVaR and the Kelly criterion, and Multi-Scale Reward Reflection leveraging $M_t = M_t^{s}+M_t^{m}+M_t^{l}$ and a trend-aligned reward. Experiments on five stocks show FinRS achieves higher cumulative returns, superior Sharpe ratios, and lower maximum drawdown than state-of-the-art LLM agents, DRL baselines, and rule-based methods. The results indicate that risk-aware, multi-timescale feedback can substantially improve robustness and profitability of real-world trading systems.

Abstract

Large language models (LLMs) have shown strong reasoning capabilities and are increasingly explored for financial trading. Existing LLM-based trading agents, however, largely focus on single-step prediction and lack integrated mechanisms for risk management, which reduces their effectiveness in volatile markets. We introduce FinRS, a risk-sensitive trading framework that combines hierarchical market analysis, dual-decision agents, and multi-timescale reward reflection to align trading actions with both return objectives and downside risk constraints. Experiments on multiple stocks and market conditions show that FinRS achieves superior profitability and stability compared to state-of-the-art methods.

FINRS: A Risk-Sensitive Trading Framework for Real Financial Markets

TL;DR

FinRS tackles risk management in LLM-based trading by embedding risk-awareness into perception, decision making, and reward. It integrates three modules: Market Perception and Analysis with hierarchical memory, Risk-Sensitive Decision Making with dual agents and dynamic position sizing using CVaR and the Kelly criterion, and Multi-Scale Reward Reflection leveraging and a trend-aligned reward. Experiments on five stocks show FinRS achieves higher cumulative returns, superior Sharpe ratios, and lower maximum drawdown than state-of-the-art LLM agents, DRL baselines, and rule-based methods. The results indicate that risk-aware, multi-timescale feedback can substantially improve robustness and profitability of real-world trading systems.

Abstract

Large language models (LLMs) have shown strong reasoning capabilities and are increasingly explored for financial trading. Existing LLM-based trading agents, however, largely focus on single-step prediction and lack integrated mechanisms for risk management, which reduces their effectiveness in volatile markets. We introduce FinRS, a risk-sensitive trading framework that combines hierarchical market analysis, dual-decision agents, and multi-timescale reward reflection to align trading actions with both return objectives and downside risk constraints. Experiments on multiple stocks and market conditions show that FinRS achieves superior profitability and stability compared to state-of-the-art methods.

Paper Structure

This paper contains 11 sections, 3 equations, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Architecture of FinRS
  4. Market Perception and Analysis Module
  5. Risk-Sensitive Decision Making
  6. Multi-Scale Reward Reflection
  7. Experiments
  8. Experimental Setup
  9. Main Results
  10. Ablation Studies
  11. Conclusion

Figures (1)

  • Figure 1: Architectural Details of FinRS: A risk-sensitive design combining perception, memory allocation, dual decision agents, and reflective updates via multi-timescale rewards.