QuantEvolver: Reinforcement Fine-Tuning for Alpha Factor Discovery
A new framework called QuantEvolver uses reinforcement fine-tuning to discover alpha factors in quantitative trading. It addresses limitations of existing LLM-based methods, such as context explosion, feedback drift, and search stagnation. The system self-evolves without relying on prompt-level loops or very large models.
Key facts
- QuantEvolver is a self-evolving alpha factor discovery framework.
- It uses reinforcement fine-tuning for optimization.
- Existing LLM-based methods suffer from context explosion and feedback drift.
- Large LLMs can cause structurally similar expressions and search stagnation.
- The framework aims to overcome these limitations.
Entities
—