RESEARCH · GLOBAL
arXiv: Risk-neutral generative networks for options pricing and density extraction
Paper presents a generative model (neural net) for option pricing that learns term structures of log-returns (location, scale, higher moments) and enforces no-arbitrage constraints during training. Enables efficient sample generation for pricing across strikes and maturities.
WHY IT MATTERS
Generative approach to options pricing is novel; enforces no-arbitrage in training loop. Relevant for BFSI desks building AI-native derivatives pricing engines. Addresses key risk: unconstrained deep learning can violate fundamental financial constraints.
Source: arXiv q-fin · 2026-05-21