LiveTradeBench: Seeking Real-World Alpha with Large Language Models

This paper describes a test bench for checking whether large language models (like ChatGPT-style systems) can actually make money in real, live trading rather than just looking smart on historical data. Think of it as a wind‑tunnel for AI trading ideas: you plug an LLM in, let it trade in real-time under controlled rules, and measure if it truly beats the market after costs and risks.