New Framework Enables LLM Trading Agents to Self-Evolve Policies

VERA Brief

AI-generated. Grounded in the article and its cited sources.

A new framework called EvolveTrade allows LLM trading agents to dynamically adapt their tool-use policies based on market data and portfolio feedback. This self-evolution aims to improve agent performance in changing market conditions by updating the agent's system prompt after each trading interval.

Key facts

• EvolveTrade enables LLM trading agents to refine operational policies without altering the core LLM.
• The framework treats the agent's system prompt as a text-parameterized policy that can be updated.
• A Policy Agent revises the prompt using decision traces and portfolio feedback.
• Experiments showed EvolveTrade often enhanced Sharpe Ratio and Cumulative Return compared to fixed policies.
• Self-evolved policies led to increased code-mediated analysis and regime-relevant computations.