Caiyue star and shanghai jiao tong university launch open-source alpha-r1 model to revolutionize financial asset allocation

Caiyue Star, a subsidiary of Jiemian Cailian, has partnered with the Antai School of Economics and Management at Shanghai Jiao Tong University to release Alpha-R1, an 8B parameter financial reasoning model. This model outperforms traditional methods and general large models in financial reasoning tasks through its 'semantic gate' reasoning technology and two-stage reinforcement learning framework. The model addresses dual challenges in asset allocation by linking financial news to investment decisions and adapting to changing market conditions. Alpha-R1 uses real-time data and semantic matching to filter investment strategies based on current market states. Its training framework combines numerical screening with semantic interviews and GRPO reinforcement learning, using real returns for continuous improvement. In tests using 2025 market data, Alpha-R1 achieved simulated annualized returns of 47.87% on the S&P 500, significantly outperforming DeepSeek-R1 and traditional methods. The model's decision-making process is traceable, with detailed analysis of its actions during the April 9, 2025, U.S. market panic. It prioritized short-term strategies while reducing long-term ones, providing clear explanations for its choices. The research team validated results with statistical tests and keyword matching experiments, confirming the model's effectiveness. The code, paper, and model are now open-sourced, with potential future applications in intelligent investment and risk monitoring.