RESEARCH · GLOBAL
arXiv: Risk-Neutral Generative Networks for option pricing and risk-neutral density extraction
Introduces a generative model for option pricing: neural nets represent term structures of log-return location, scale, and higher moments, with no-arbitrage constraints enforced during training. Enables efficient option valuation across strikes and maturities.
WHY IT MATTERS
Offers an alternative to classical Black-Scholes for derivatives pricing; BFSI quant teams should backtest this approach on equity options books to assess whether generative models improve mark accuracy and reduce tail-risk mispricing.
Source: arXiv · 2026-05-21