LiveOption: An LLM Agent Evaluation Framework for Option Trading with Nonlinear Payoffs

Researchers have released the LiveOption option trading agent evaluation framework, which models option trading as a structured sequential decision-making problem under realistic execution and capital constraints, accompanied by a standardized interaction protocol and a reproducible experimental environment. The framework includes three task suites: portfolio hedging, event-driven earnings trading, and 0DTE intraday trading, along with a layered metric system that evaluates agent performance across four dimensions: action validity, decision quality, risk characteristics, and outcome performance. Experiments show that current mainstream LLM agents struggle to achieve competitive returns in most scenarios; the project homepage is available through the paper page.