Forecasting implied volatility surface with generative diffusion models

Paper: arXiv 2511.07571

Authors: Chen Jin, Ankush Agarwal

Abstract

We introduce a conditional Denoising Diffusion Probabilistic Model (DDPM) for generating arbitrage-free implied volatility (IV) surfaces, offering a more stable and accurate alternative to existing GAN-based approaches. To capture the path-dependent nature of volatility dynamics, our model is conditioned on a rich set of market variables, including exponential weighted moving averages (EWMAs) of historical surfaces, returns and squared returns of underlying asset, and scalar risk indicators like VIX. Empirical results demonstrate our model significantly outperforms leading GAN-based models in capturing the stylized facts of IV dynamics. A key challenge is that historical data often contains small arbitrage opportunities in the earlier dataset for training, which conflicts with the goal of generating arbitrage-free surfaces. We address this by incorporating a standard arbitrage penalty into the loss function, but apply it using a novel, parameter-free weighting scheme based on the signal-to-noise ratio (SNR) that dynamically adjusts the penalty’s strength across the diffusion process. We also show a formal analysis of this trade-off and provide a proof of convergence showing that the penalty introduces a small, controllable bias that steers the model toward the manifold of arbitrage-free surfaces while ensuring the generated distribution remains close to the real-world data.

Complexity vs Empirical Score

• Math Complexity: 8.5/10
• Empirical Rigor: 7.0/10
• Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper features advanced mathematics, including diffusion processes, SDEs, and a formal convergence proof, while also demonstrating strong empirical rigor with a comparative benchmark (GAN), backtest-ready code, and quantitative metrics like MAPE and confidence intervals.

Research Flowchart

  flowchart TD
  A["Research Goal: <br>Forecast IV Surface<br>Arbitrage-Free with DDPM"] --> B{"Data & Inputs"}
  B --> B1["Market Variables<br>EWMA of IV, Returns,<br>Squared Returns, VIX"]
  B --> B2["Historical IV Surfaces<br>Contains Minor Arbitrage"]
  
  B1 & B2 --> C["Conditional DDPM Framework"]
  
  C --> D{"Arbitrage Penalty<br>in Loss Function"}
  D --> D1["Parameter-Free Weighting<br>Based on Signal-to-Noise Ratio"]
  
  C --> E["Computational Training<br>Stochastic Gradient Descent"]
  E --> F
  
  subgraph F ["Key Methodological Innovation"]
      D1 --> F1["Dynamic Steering<br>Toward Arbitrage-Free Manifold"]
  end
  
  F1 --> G["Outcomes & Findings"]
  G --> G1["Significantly Outperforms<br>GAN-based Models"]
  G --> G2["Stylized Facts<br>of IV Dynamics Captured"]
  G --> G3["Formal Proof:<br>Bias-Variance Trade-off"]

Cited References (31)

1. Score-Based Generative Modeling through Stochastic Differential Equations (2021) — link
2. Reverse-time diffusion equation models (1982) — link
3. A Connection Between Score Matching and Denoising Autoencoders (2011) — link
4. Sigmoid-weighted linear units for neural network function approximation in reinforcement learning (2018) — link
5. Conditional score-based diffusion models for Bayesian inference in infinite dimensions (2023) — link
6. Efficient Diffusion Training via Min-SNR Weighting Strategy (2023) — link
7. Conditional Diffusion Models are Minimax-Optimal and Manifold-Adaptive for Conditional Distribution Estimation (2024) — link
8. Comparing Predictive Accuracy (1995) — link

1 of 31 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.