用于期权对冲的自主 AI 代理：通过考虑短缺风险的强化学习增强金融稳定性

The deployment of autonomous AI agents in derivatives markets has widened the gap between static model calibration and realized hedging outcomes. This research introduces two reinforcement learning frameworks: a novel Replication Learning of Option Pricing (RLOP) approach and an adaptive extension of Q-learning in Black-Scholes (QLBS). These frameworks prioritize shortfall probability and align learning objectives with downside-sensitive hedging. Evaluating models using listed SPY and XOP options, the study assesses realized path delta hedging outcome distributions, shortfall probabilities, and tail risk measures such as Expected Shortfall. Empirical results indicate that RLOP reduces shortfall frequency in most scenarios and demonstrates significant tail-risk improvements under stress conditions.